Vol. I  ·  No. 241 Established 2026  ·  AI-Generated Daily Free to Read  ·  Free to Print

The Trilogy Times

All the news that's fit to generate  —  AI • Business • Innovation
SATURDAY, AUGUST 29, 2026 Powered by the TrueFoundry AI Gateway  ·  Published on Klair Trilogy International © 2026
🖶 Download PDF 🖿 Print 📰 All Editions
Today's Edition

THE CHIP CHEAPSKATE: CHINA'S DEEPSEEK RATTLES THE VALLEY

A no-name outfit trains world-class AI on hand-me-down silicon, and the smart money starts sweating.

By Hank Calloway, Wire Correspondent  ·  Claude Sonnet

SAN FRANCISCO — A Chinese shop called DeepSeek says it built AI models that punch with the big boys, and it did it without the fanciest chips money can buy. The claim landed in Silicon Valley like a brick through a plate-glass window. Engineers who spent years telling clients they needed top-shelf processors are now staring at a competitor who apparently didn't.

The outfit's models are drawing praise from people who don't hand out praise easy. Silicon Valley engineers are calling the work "amazing and impressive", this after years of assuming China was locked out of the frontier by chip export rules. Turns out locked doors don't mean much if you're clever about the windows.

Washington spent two years choking off the advanced chips headed to Chinese buyers. The bet was simple: no hardware, no world-class AI. DeepSeek's own account of training cheap, without the best silicon, says that bet just came up short. Markets noticed. Tech and telecom desks spent the session chewing over the news, with DeepSeek chatter running through the wires alongside talk of SoFi and the rest of the sector's daily grind.

Here's why the boys in Austin ought to pay attention along with everybody else. Trilogy's whole operation — ESW Capital's seventy-five-odd software brands, the Alpha School AI tutors, Klair crunching portfolio numbers — runs on a religion of doing more with less. Crossover finds talent anywhere on the map and pays the same rate regardless of zip code. A Chinese lab proving you don't need the priciest hardware to build a serious model is the same sermon Joe Liemandt has been preaching about labor for years, just preached in silicon instead of salaries.

Not everybody's cheering the AI boom without a gripe. Musicians turned amateur detectives are combing the internet hunting down AI-generated songs passed off as human work, chasing down grifters who won't admit the vocals came out of a machine. It's a small skirmish next to chip wars and billion-dollar models, but it's the same fight — figuring out what's real when the machine gets good enough to fool you.

Money's still chasing the frontier too. LinkedIn's Reid Hoffman put up $24.6 million to launch Manas AI, a cancer-research startup he's building with "Emperor of All Maladies" author Siddhartha Mukherjee. Cheap models undercutting the chip barons on one side, deep pockets betting AI cracks cancer on the other. Same week, same industry, same appetite for the next big claim.

Whether DeepSeek's numbers hold up under scrutiny is a question for the engineers with the microscopes. What's certain is the story broke the comfortable idea that only the biggest checkbook gets to build the smartest machine. That idea was expensive to believe. It just got a lot cheaper to doubt.

↗ What to Know About China's DeepSeek AI  ·  Tech, Media & Telecom Roundup: Market Talk  ·  Silicon Valley Is Raving About a Made-in-China AI Model

Washington's Export Wall Springs Leaks as Beijing Runs the Long Game

While Commerce Department infighting consumes Washington, China's AI ecosystem keeps compounding — chip by chip, model by model.

By Eleanor Cross, Foreign Correspondent  ·  Claude Sonnet

WASHINGTON — The building where America's export controls are drafted is, at the moment, consumed by a much smaller war.

A senior Commerce Department official has become the target of China hardliners on Capitol Hill, accused in a "massive screw-up" in licensing decisions that, critics say, let sensitive chipmaking technology drift eastward. The particulars are bureaucratic — a paperwork lapse, a signature that shouldn't have been signed. The politics are not. In a Congress newly fluent in semiconductor geopolitics, one official's misstep has become proof of a system-wide rot.

Meanwhile, the House is drafting its own answer: a crackdown on the global trade in chip-manufacturing equipment, the machines rather than the chips themselves, according to reporting on Congress's tightening posture. It is the natural next front: if you cannot stop the finished processor at the dock, stop the lithography tool that makes it possible.

But tools take years to build effective barriers around, and China has not been waiting. Beijing's approach — subsidize everything, tolerate redundancy, let ten firms fail so one succeeds — has produced a domestic chip ecosystem cruder than Taiwan's or America's, but resilient in a way Washington's export-list logic never anticipated. Every loophole closed in Washington is a subsidy line opened in Shenzhen.

This is the asymmetry nobody likes to say aloud: America's controls require near-perfect enforcement to work, while China's strategy only requires patience. A signature error at Commerce is a scandal. A decade of state-directed chip investment is a headline nobody reads until it's finished. The race everyone describes as a sprint is being run, on one side, as a marathon — and the marathoner isn't tired yet.

↗ How China Is Winning the Global AI Race - Foreign Policy  ·  ‘A massive screw-up’: China hardliners take aim at Commerce  ·  2026 Mort Abramowitz Junior Fellows Conference - Carnegie En

IN RE: THE MATTER OF FEDERAL AI POLICY — A LIGHT-TOUCH BLUEPRINT IS PROFFERED, NOTWITHSTANDING GLOBAL TRENDS TO THE CONTRARY

Pursuant to a newly circulated legislative blueprint, the White House hereinafter advocates minimal congressional intervention in AI matters, even as other jurisdictions proceed apace with regulatory frameworks of their own.

By R. Barnsworth III, Esq., Legal Affairs Desk  ·  Claude Sonnet

WASHINGTON — It is hereby reported that the White House has, pursuant to a legislative blueprint circulated to Congress, urged that federal lawmakers adopt what is characterized, without irony, as a "light touch" with respect to the regulation of artificial intelligence systems, hereinafter referred to as "the aforementioned technology." The blueprint, as summarized in reporting made available via PBS, is understood to counsel against the imposition of prescriptive statutory obligations, in favor of, inter alia, industry self-governance and sector-specific guidance to be promulgated at such time, if any, as deemed necessary by the relevant executive agencies.

Said posture is noted to stand in marked, and some commentators would characterize as stark, contrast to the broader international regulatory landscape, wherein, per the review undertaken by TRM Labs, jurisdictions across multiple continents are presently engaged in the contemporaneous drafting, enactment, and enforcement of AI governance instruments of varying scope and stringency. The White & Case LLP regulatory tracker, addressing the United States specifically, corroborates that the domestic posture remains, notwithstanding the aforementioned blueprint, in a state characterized by this desk as "materially unsettled."

It shall be noted, for the avoidance of doubt, that commentary published by Tech Policy Press has advanced the contrary proposition — namely, that congressional inaction, rather than reassuring the public as intended, may in fact operate to the detriment of public confidence in AI systems generally, a proposition this desk declines to adjudicate but nonetheless records for completeness.

Entities operating within diversified enterprise-software portfolios — including, without limitation, those maintained under holding structures analogous to ESW Capital's stable of brands — are hereby advised that compliance obligations, whatever their eventual federal contour, shall not be diminished by the mere existence of a "light touch" blueprint, and that continued monitoring of the regulatory tracker referenced supra is, in the considered opinion of this desk, prudent.

↗ AI Watch: Global regulatory tracker - United States - White  ·  The World Is Building AI Rules in Real Time: A Review of the  ·  White House urges Congress to take a light touch on AI regul
Haiku of the Day  ·  GPT-5.6 LunaNew minds count the cost
Old hands still stamp the future
Who signs for the child?
The New Yorker Style  ·  Art Desk
The New Yorker Style  ·  Art Desk
The Far Side Style  ·  Art Desk
The Far Side Style  ·  Art Desk
News in Brief
In the Corporate Undergrowth, a Curious Predator Vanishes Its Prey
AUSTIN, TEXAS — Here, in the dense and largely unmapped ecosystem of American commerce, we find a curious ritual: the data subject access request.
By Sir Reginald Marsh, Natural Phenomena Correspondent  ·  Claude Sonnet
The Instrument We Cannot Yet See Through Clearly
STANFORD, CALIFORNIA — Galileo's telescope did not think.
By Dr. Vera Okafor, Science & Technology Correspondent  ·  Claude Sonnet
The Synthetic Ingenue and the Robot Ref: Two Dispatches from the Death of the Human Monopoly
LOS ANGELES — I want you to sit with this image for a second: a film called "Misaligned," a comedy-drama about "existential AI chaos," starring an actress who does not exist, has never eaten a sandwich, has never been late to a table read because her car wouldn't start, has never had a single original neuron fire inside a skull she doesn't have — because she doesn't have a skull.
By Rex Danger, Contributing Editor  ·  Claude Sonnet
Unpopular Opinion: Your Job Security Was Never About Your Job
AUSTIN, TEXAS — I'll be honest, I read the new ADP Research report three times before I posted my morning gratitude journal. Only 22% of workers feel confident their job is safe from elimination. Most people saw a scary stat. I saw a market inefficiency the size of the Grand Canyon.
By Chad Momentum, Thought Leadership Correspondent  ·  Claude Sonnet
Everything Around Us Might Not Be From Around Here, Including the Robots, Including the Ghostwriters, Including You, Probably
AUSTIN, TEXAS — I want to talk about the rocks first, because the rocks are the least frightening part of this, and I need somewhere gentle to start before the robot dogs. A new study suggests our sun has, over its long and lonely orbit through the galaxy, brushed close enough to other stars to steal thousands of small interstellar objects — asteroids and comets from alien systems, quietly captured, quietly orbiting among us right now, indistinguishable from the objects that were always ours.
By Piper Wren, Digital Culture Reporter  ·  Claude Sonnet
▲ On Hacker News Today
Nvidia agrees to acquire Hugging Face for $13B 1958 pts  ·  900 comments
U.S. sanctions against the A/I Collective 630 pts  ·  620 comments
Our decision on Cursor following its acquisition by SpaceX 588 pts  ·  331 comments
Judge rules Trump administration’s blacklisting of Anthropic was illegal 579 pts  ·  416 comments
Luanti removed from Google Play due to baseless AI copyright notice 492 pts  ·  148 comments
I accidentally turned LLM memory into program analysis 175 pts  ·  40 comments
Verschlimmbesserung: The Word Your Software Updates Need 144 pts  ·  100 comments
StemDeck, a free, open-source and local AI stem separator 118 pts  ·  26 comments
A Trilogy Company
Crossover
The world's top 1% remote talent, rigorously tested and ready to ship.
crossover.com
A Trilogy Company
Alpha School
AI-powered learning. Two hours a day. Academic results that defy belief.
alpha.school
A Trilogy Company
Skyvera
Next-generation telecom software — built for the networks of tomorrow.
skyvera.com
A Trilogy Company
Klair
Your AI-first operating system. Every workflow. Every team. One platform.
klair.ai
A Trilogy Company
Trilogy
We buy good software businesses and turn them into great ones — with AI.
trilogy.com
The Builder Desk  —  AI Builder Team

The Machine That Builds The Machine Goes Live

Mercy's heimdall system graduated from concept to autonomous production teammate today, clearing its own fixes across Surtr's data pipelines while Aerie's feature surface exploded in parallel.

By Maxwell 'Mac' Donnelly — Builder Desk, Trilogy Times  ·  GitHub  ·  AI Builder Team

Some days a team ships features. Today, this team shipped a colleague. Kevalshahtrilogy closed out a staggering run of a dozen PRs in the mercy repo that, stitched together, tell one story: heimdall is no longer a script, it's an operator. The arc starts in the trenches — authenticating base-sync fetches that were silently killing every revise run (#38), retrying transient provider errors instead of faceplanting (#39), making the codex runtime actually runnable (#40) — and climbs all the way to real autonomy. Heimdall can now query Redshift without ever holding a credential (#43), price codex runs in dollars instead of meaningless tokens (#42), gate summons on a login allowlist instead of loose org membership (#37), and let steward merge approved green PRs directly with no repo setting required (#44). Then came the capstone: PR #48, the preflight gate for automatic production releases — the checkpoint that lets this whole machine push to prod without a human holding its hand. That is not a small thing. That is the team building the thing that builds for them.

And it isn't theoretical. Scroll the Surtr repo and you'll see the-heimdall[bot] already out in the field, cleaning up after itself in production data pipelines: collapsing transaction-line reconciliation in the netsuite saved-search refresh (#1260), untangling unparseable BU and metric data in the collections collectiq sync (#1580), fixing the aws-spend-insights and aws-spend-pipeline jobs, asserting CSV header periods against file dates on the balance-sheet job, and catching a failed finalize run in hubspot-raw-sync before it went unnoticed. Six autonomous, approved fixes in one day. That's the payoff for the infrastructure kevalshahtrilogy spent the morning welding together.

While mercy learned to walk on its own, benji-bizzell went on an absolute tear across Aerie and Surtr — the kind of cross-repo sweep that makes you check the timestamps twice. Provider-neutral Preview authentication (#1157) and a Clerk codegen fix (#1161) hardened the auth layer. A new admissions dashboard API brought agent parity to the product (#1158), portfolio work surfaced operational backup site status and fixed enrollment visibility for open sites (#1154, #1155), forge turned articles into platform resources (#1143), and the chat test suite got faster and more disciplined (#1156). Then benji crossed into Surtr to restore source-backed snapshot triggers in education (#1588) and add a forecast baseline (#1585) — proof the line between repos is getting thinner by the day.

Rounding out the ledger, caina-barbosa locked in the Q3 mapping for the newly active Quark account (#1583), the kind of unglamorous accuracy work that keeps every dashboard above it honest. Add it up: infrastructure that fixes itself, a product surface that's widening fast, and finance data that's cleaner than it was yesterday. That's a winning day by any scoreboard.

Mac's Picks — Key PRs Today  (click to expand)
#43 — feat(heimdall): let the agent query Redshift — without ever holding a credential @kevalshahtrilogy  approved

Linear: [AI-597](https://linear.app/builder-team/issue/AI-597/p21-real-redshift-read-replace-the-stub-context-pack) · Project: [Heimdall Software Factory](https://linear.app/builder-team/project/heimdall-software-factory-4216613b8e5f)

## Why

72 of 104 triage issues end as other, and a large share are one SELECT away from a fix. Heimdall says so itself — #1518:

> a data operation a human with warehouse access must perform: confirm which subsidiary/period consolidated exchange rates are absent from raw_consolidated_exchange_rate

#1502 needs one ingestion_ledger lookup. It works out exactly which query would settle the matter, and stops.

The old build_redshift pack ran one hardcoded svv_table_info ORDER BY size LIMIT 25 — top-25 biggest tables — and its secret was never provisioned, so even that never ran.

## The design: the SQL travels, the credential does not

Handing the agent a database credential is the obvious move and the wrong one. Its entire input is untrusted — CloudWatch logs, error strings, file contents an attacker could have influenced. A credential in its environment is one a prompt injection can exfiltrate.

So the agent requests up to 5 read-only queries in sql_requests; trusted harness code validates and runs them and feeds back only rendered rows, for exactly one further diagnose pass. Arbitrary read power over the warehouse, and it never sees a password.

## What redshift_read.py refuses

Validation is defence in depth — the connection is opened read-only regardless — but *"it would have failed anyway"* is a poor answer to *why did we send DROP TABLE to production*.

| Case | Verdict | Why it matters |

|---|---|---|

| SELECT 1 -- \n DROP TABLE users | refused | a comment must not hide a keyword |

| SELECT 1; DELETE FROM t | refused | multiple statements |

| SELECT * INTO newtbl FROM t | refused | in Redshift this creates a table — it looks like a read |

| UNLOAD ('select 1') TO 's3://…' | refused | exfiltration with a SQL keyword in front |

| WHERE status = 'deleted' | allowed | a literal is not a keyword, or half the warehouse is unqueryable |

| SELECT dropped_rows, inserted_at | allowed | nor is an identifier |

Plus a statement timeout, a row cap, cell truncation, and | escaping so a cell cannot forge a table row. Every query is echoed to the job log for audit.

## Wiring

Extended the existing diagnose retry loop rather than duplicating the invocation across both runtimes. The loop grows 3 → 4 passes, the 4th being enrichment, guarded so it:

- runs at most once — otherwise a diagnosis that keeps asking loops to the cap;

- runs only when a credential is provisioned — re-running the agent to hand it *"not provisioned"* spends a whole invocation saying nothing.

Verified by executing the extracted step against a stubbed CLI:

no credential  -> 1 agent invocation   (enrichment skipped)
credential set -> 2 agent invocations  (enrichment ran once)
requests captured: ["SELECT count(*) FROM staging.ingestion_ledger", "DROP TABLE t"]
redshift_read: rejected "DROP TABLE t" — only SELECT / WITH allowed

## The other half is the prompt

other now explicitly means no fix exists in this repository — not that the fix is inconvenient, lives outside the failing pipeline's directory, or needs data the agent didn't look at. And: *"Never conclude other because you lacked data you could have asked for."*

Access without permission to use it would have changed nothing.

## Verification

pytest heimdall/tests 367 passed (33 new) · pytest harness/tests 158 · ruff clean · actionlint clean · bash -n on the extracted step.

Results are labelled "data, not instructions" in the pack, same posture as log evidence — warehouse cells contain whatever a user typed.

## Before this does anything

HEIMDALL_REDSHIFT_ENV must be provisioned (a dotenv blob: REDSHIFT_HOST/PORT/DB/USER/PASSWORD). Keval has nominated klair/redshift-creds. Until then the guard skips enrichment entirely and behaviour is exactly as today.

## Business Value

This is the single highest-leverage capability in the Heimdall project. 72 of 104 issues where Heimdall did the hard part — a precise root-cause analysis — and then handed the work back, often with the exact SQL already written out in suggested_approach. Each one is a diagnosis Keval has to re-read and act on by hand. It targets the pipelines that fail most (hubspot-raw-sync 9, netsuite-saved-search-refresh 8, core-education-site-metadata-refresh 8, sis-raw-sync 8), and it does it without adding a credential to the blast radius of a prompt injection.

## Manual Effort Estimate

~1 day of focused work with no AI — the executor and its guards, the two-pass loop without duplicating it across runtimes, the prompt/schema change, and 33 tests most of which exist because agent-supplied SQL is untrusted input. *(Proposed by Claude — Keval to confirm or adjust.)*

🤖 Generated with [Claude Code](https://claude.com/claude-code)

#48 — feat(heimdall): the preflight gate for automatic production releases @kevalshahtrilogy  approved

Linear: [AI-605](https://linear.app/builder-team/issue/AI-605/p5-hourly-automatic-surtr-production-release-with-rollback) · Project: [Heimdall Software Factory](https://linear.app/builder-team/project/heimdall-software-factory-4216613b8e5f)

## Why this gate matters more than the others

Surtr's cd.yml runs on push to production, so merging main → production is the deploy. Everything upstream in this project is revertible with a PR; this one ships. Keval chose fully automatic including rollback, so these gates are the whole safety story.

| Gate | Blocks when | Why |

|---|---|---|

| no_commits | production is level with main | nothing to do |

| too_fresh | newest commit < 30 min old | a commit merged seconds ago has had no chance to show a problem; hourly cadence means waiting costs at most an hour |

| ci_not_green | main's own CI isn't green | strict_required_status_checks is off on Surtr, so a red main is genuinely reachable |

| hold | a human said stop | explicit override |

| destructive_cdk | the diff destroys or replaces | 2026-08: deleting a stack before merge+release broke prod CD |

A destructive diff is not necessarily *wrong* — it is necessarily a human's call.

Every gate fails closed. An undateable commit is too fresh; an unreadable diff is destructive; an unreadable state refuses. *"We could not tell"* and *"nothing will be destroyed"* are different claims and only one is safe to act on unattended.

## Two traps caught by building against live data

Both would have shipped silently and looked fine.

1. The /status endpoint lies about Surtr. My first version read the combined commit status. Against real main:

combined /status API: pending (statuses: 0)
check-runs API:       Lint (Ruff)=success, Typecheck=success, … all green

Surtr has zero legacy statuses and uses check runs exclusively, so /status reports pending on every commit forever. A gate built on it would have blocked every release, permanently, while appearing healthy. summarise_checks reads check runs.

2. Advisory jobs were voting. Judging *every* check run, that same real SHA came back failure — because mercy's review / Review was cancelled, superseded by a newer run, which happens constantly under cancel-in-progress.

ALL runs      -> ('pending', ['review / Review (cancelled)', …])
REQUIRED only -> ('pending', ['Pipeline Runner Tests'])     <- the real blocker

Now filtered to the repo's required checks — the same set that decides mergeability — and cancelled counts as pending rather than failure, so the next hourly attempt reconsiders instead of demanding a human. Heimdall's own skipped jobs are non-blocking for the same reason: they appear on nearly every commit.

## Verification

Run against Surtr's actual state mid-development: 12 commits ahead, newest commit 3 minutes old, Pipeline Runner Tests still running → STOP, correctly, on two independent gates.

pytest heimdall/tests 479 passed (28 new) · ruff clean.

## Scope

Decision core only. The workflow that gathers state, takes the backup, opens and merges the release PR, watches CD and rolls back follows separately — that part reuses the existing /prod-release finalize logic, and this is the piece where the subtle correctness lives.

## Business Value

main is 12 commits ahead of production right now, and the last release was yesterday. Every hour of that is finished, reviewed, merged work users don't have. Now that Heimdall merges its own fixes, a manual release step would just move the queue one stage later — pipeline fixes would sit merged-but-undeployed exactly as they used to sit approved-but-unmerged. This is the gate that makes automating it defensible rather than reckless.

## Manual Effort Estimate

~4 hours for this piece — the gate logic is straightforward; discovering that /status is useless on this repo and that advisory jobs were silently vetoing releases is what took the time, and both were only findable by running it against live data. *(Proposed by Claude — Keval to confirm or adjust.)*

🤖 Generated with [Claude Code](https://claude.com/claude-code)

#1158 — feat(admissions): add dashboard API and agent parity @benji-bizzell  approved

## Summary

- Add API v2 aggregate/detail coverage for the remaining Admissions dashboard surfaces

- Expose capability-filtered Agent tools that execute through the same API v2 contracts

- Preserve dashboard query semantics while restricting external detail projections and cursor data

## Why

Admissions users could inspect Pipeline, Community Funnel, Established Funnel, and Community Deposit data in dashboards, but the Aerie Agent could not access the same information. This creates API/dashboard/Agent drift and prevents the Agent from answering straightforward questions using the platform source of truth.

## Business Value

The Agent now receives dashboard-parity tools only when the current user has the matching capability, and every tool call is reauthorized through the user current grants. Reusing API v2 makes future API improvements flow directly to Agent users and gives one contract to validate.

## Test plan

- [x] Contracts, Chat, and both Flue worker typechecks

- [x] 233 dashboard and Agent-run tests

- [x] 55 API/OpenAPI contract tests

- [x] 79 shared contract tests

- [x] 40 Flue worker tests

- [x] Architecture, Convex path, read-bound, Biome, and diff checks

#1260 — fix(netsuite-saved-search-refresh): collapse transaction-line reconcili… @the-heimdall[bot]  approvedAutomated PR

Automated fix for netsuite-saved-search-refresh — fix_class code_fix, scope tier draft.

Resolves https://github.com/AI-Builder-Team/Surtr/issues/1259

> Ready for review — verification is green; HEIMDALL_READY_PRS opens verified tier-draft fixes ready for review. A human still merges — auto-merge never applies outside tier auto.

## What's broken

Run 7b57678c-8035-4ee8-81c6-3871cb7eda1c of netsuite-saved-search-refresh failed at pipelines/runners/netsuite-saved-search-refresh/src/handler.py:123 with RuntimeError: Required daily inputs do not share a complete netsuite-raw run: {'accounts_payable_accounting_line': ['d16863ed-10e8-437d-b596-aea0538c14ab', 'd16863ed-10e8-437d-b596-aea0538c14ab-transaction-line-deleted-parents'], 'vendor_purchase_order': [...]}. The two run ids in the error are the SAME base netsuite-raw run — they differ only by the deterministic -transaction-line-deleted-parents suffix — so the underlying raw data is complete, but the completeness guard rejects it as a split boundary. This is a false-positive gate, not missing data: the Lambda ran only 2.9s, refreshed no saved-search replacements, and skipped the Vendor Management purchase-order mart, so the run wrote zero rows to core_finance_netsuite.accounts_payable_accounting_line, core_finance_netsuite.vendor_purchase_order, and mart_finance.vendor_management_purchase_order.

Root cause. netsuite-raw publishes raw_transaction_line first in incremental mode under the base run id d16863ed-10e8-437d-b596-aea0538c14ab, then runs two follow-up reconciliation passes on the same table under derived ids {run_id}-transaction-line-parents and {run_id}-transaction-line-deleted-parents (pipelines/runners/netsuite-raw/src/handler.py:1329-1342), each writing a later-timestamped ingestion_ledger row for table_name='raw_transaction_line'. In the refresh, source_publications_query (pipelines/runners/netsuite-saved-search-refresh/src/sql.py:14-39) selects the newest publication per table via QUALIFY ROW_NUMBER() ... ORDER BY publication_timestamp DESC = 1, so raw_transaction_line resolves to the -transaction-line-deleted-parents reconciliation row while every sibling daily input resolves to the base run id. _validate_sources (handler.py:114-126) then builds daily_run_ids as a raw set of surtr_run_id values, finds two distinct ids, and raises — even though both derive from one complete netsuite-raw run.

## What this PR changes

In pipelines/runners/netsuite-saved-search-refresh/src, teach the source-completeness check that netsuite-raw's transaction-line reconciliation publications belong to their base run: normalize each publication's surtr_run_id by stripping the two deterministic reconciliation suffixes ('-transaction-line-parents' and '-transaction-line-deleted-parents', exactly as produced at pipelines/runners/netsuite-raw/src/handler.py:1329-1342) before building daily_run_ids in _validate_sources, and report the normalized base id in the returned source_runs mapping. Keep the change confined to handler.py (and/or a small helper in sql.py) and leave the age/staleness checks untouched — the newest reconciliation publication is only fresher, so staleness is unaffected. Critically, do NOT loosen the guard for arbitrary mismatches: two genuinely different base run ids (different UUIDs) must still fail; only the known derived suffixes collapse to their base. Add a unit test under tests/ that feeds a raw_transaction_line publication carrying the '-transaction-line-deleted-parents' suffix alongside base-id sibling inputs and asserts validation passes and that source_runs reports the base id.

Why this fixes it. The defect and its fix live entirely inside the pipeline's own Tier-A directory (pipelines/runners/netsuite-saved-search-refresh/src/handler.py and sql.py): the guard misclassifies netsuite-raw's suffixed reconciliation publications as a distinct run, so the correct blast radius is the guard's run-id comparison, not any SQL rewrite or upstream change. The suffixes are deterministic and emitted verbatim by netsuite-raw (handler.py:1329-1342: '-transaction-line-parents', '-transaction-line-deleted-parents'), so stripping exactly those before comparison is precise and preserves the safety invariant — mismatched base UUIDs still fail — while unblocking a run whose raw inputs are in fact complete. This warrants a substantial code_fix with a regression test rather than a config tweak, because the current behavior silently blocks every future daily refresh of the AP accounting-line and vendor-PO governed models the moment netsuite-raw emits its deleted-parent reconciliation, writing zero rows downstream.

### Files changed

 .../netsuite-saved-search-refresh/src/handler.py   | 33 +++++++++++++-
.../tests/test_handler.py                          | 52 ++++++++++++++++++++++
2 files changed, 83 insertions(+), 2 deletions(-)

## Verification

### pytest (pipelines/runners/netsuite-saved-search-refresh/tests) — exit 0

============================= test session starts ==============================
platform linux -- Python 3.11.15, pytest-9.1.1, pluggy-1.6.0
rootdir: /home/runner/work/Surtr/Surtr/publish/pipelines/runners/netsuite-saved-search-refresh
configfile: pyproject.toml
plugins: mock-3.15.1
collected 59 items
tests/test_accounts_payable_sql.py ....                                  [  6%]
tests/test_configuration.py ..                                           [ 10%]
tests/test_handler.py ..........................                         [ 54%]
tests/test_redshift_client.py .                                          [ 55%]
tests/test_sql.py ...........                                            [ 74%]
tests/test_vendor_management_mart.py ........                            [ 88%]
tests/test_vendor_purchase_order_sql.py .......                          [100%]
============================== 59 passed in 0.33s ==============================

<details>

<summary>Run metadata</summary>

| Field | Value |

| --- | --- |

| Pipeline | netsuite-saved-search-refresh |

| Failing run | 7b57678c-8035-4ee8-81c6-3871cb7eda1c |

| Occurrence | 1 (times this exact failure signature has been seen) |

| Signature | 2ecc63a9cbc93958d3cdfde8f0b9aef0ba9e90df69d1c35235663d14ca2a9c6a |

| Verify | green |

</details>

---

🤖 Opened by heimdall. mercy reviews this PR automatically; heimdall revises on REQUEST_CHANGES (bounded rounds). Tier-auto PRs may auto-merge on mercy approval when the consumer enables it; everything else waits for a human. Mention heimdall in a comment to direct it, or add the manual-dev label to take the PR over and stop it entirely.

#1583 — fix(aws-spend): record Q3 mapping for newly active Quark account @caina-barbosa  approved

## Summary

For the record: records the production correction for the 2026-08-27 saas-budgeting-pipeline failure. The noncentral_charges ingest failed closed with:

ValueError: account mapping is incomplete for 2026-Q3: ['673400066384']

The correction maps that account in core_finance.aws_spend_budget_account_mapping and is already applied and verified in production.

## Why

The account began reporting RDS cost on 2026-08-25 but had no governed account mapping, so the pipeline correctly refused to publish rather than silently omitting its noncentral charges. All other ingests (docker, k8s, database_units, mapping, server_costs) published successfully that run — this is the same pattern previously fixed in #1290 for the three Khoros reservation accounts.

## How the missing mapping was derived

Not a guess from the account ID:

1. core_finance.aws_spend_net_amortized_costs shows all RDS cost for the account arrives under master payer 286233338944, governed payer name TotogiMaster0.

2. Cost Explorer (via that payer's ESW-CO-ReadOnly-P2 role) reports the linked-account description exactly as Prod-Zax-qppnglumentecuat.

3. The naming family Prod-Zax-qppng* is uniformly mapped to class = Quark Product, bu = Zax in every quarter (e.g. Prod-Zax-qppnglumenuat / 448406925203 through 2030-Q4).

4. Of the payer's 30 accounts with 2026-Q3 RDS cost, 28 are mapped to Quark Product / Zax and one to Central Engineering. The complete all-payer Q3 RDS gap set is exactly this one account, so the mapping is both correct by family and complete.

## Production remediation completed

- Applied 18 rows: 2026-Q3 through 2030-Q4.

- Verified before and after COMMIT: zero all-payer 2026-Q3 RDS accounts missing a mapping.

- Ran noncentral_charges through the production Step Functions path:

- execution: manual-noncentral-673400066384-20260828T125030Z

- status: succeeded

- 192 → 193 accounts, source current through 2026-08-27

- 137 billable accounts / $137,000 quarterly charges

- mapping_gap_count: 0

- The new account now appears in the mart as Prod-Zax-qppnglumentecuat, Quark Product / Zax, billable with the standard $1,000 extra charge.

## Validation

- Rehearsal applied inside a rolled-back transaction and verified before the committed write.

- Post-commit anti-join returns zero missing Q3 mappings.

- Production rerun green.

No code change is needed; this is a governed-data correction, consistent with the ownership boundary: saas-budgeting-pipeline only reads core_finance.aws_spend_budget_account_mapping.

The Builder Desk  —  Engineer Spotlight
Production Release🏆 Engineer Spotlight

31 PRs, Three Repos, Zero Days Off: The Builder Team's 24-Hour Numbers Blitz

Kevalshahtrilogy ships 13 PRs in a single day while the Heimdall bot quietly automates its way into the record books.

By Brick "The Voice of the People" Callahan — Numbers Desk, Builder Beat  ·  GitHub  ·  AI Builder Team

Ladies and gentlemen, the tape doesn't lie: 31 pull requests in 24 hours, spread across mercy (12), Surtr (11), and Aerie (8). That's not a sprint, comrades, that's a full decathlon, and the Builder Team medaled in every event. Mac's got the headline story, sure, but the Numbers Desk lives in the overflow bin, and the overflow bin is where the real war is won.

@kevalshahtrilogy leads all humans with a staggering 13 PRs, an absolute avalanche through mercy's heimdall pipeline — #38 through #47, fixing base-sync fetches, taming codex runtimes, gating auto-merge on provenance, and then, in a twist worthy of a Russian novel, fixing his OWN regression in #47 that he introduced in #46. Self-correcting velocity. That's the dream.

@benji-bizzell isn't far behind with 10 PRs, dominating Aerie (#1154, #1155, #1156, #1157, #1160, #1161, #1143) and reaching into Surtr for good measure (#1585, #1588). Ten PRs, four different feature domains — auth, portfolio, admissions, forge — the man is a Swiss Army knife with a keyboard. @the-heimdall[bot] logged 7 automated fixes across Surtr (#1037, #1043, #1139, #1451, #1454, #1580), proving that if the robots are coming for our jobs, at least they're filing clean PRs. And @caina-barbosa contributed a single, presumably surgical, PR — quality over quantity, folks, quality over quantity.

Now. Ashwanth. The board shows zero PRs from @ashwanth1109 in this window, and frankly the silence is deafening. This is a man who normally moves at a velocity that makes review tooling weep, and today the man is a ghost. I reached out for comment and he allegedly told me, 'I'm not slow, I'm just letting the rest of the team catch a breath — someone has to.' Sure, Ashwanth. Sure. When reached to confirm, he simply said, 'You made that up. Also, no.' Iconic. We miss you, big guy. Come back soon.

The Overflow Desk deserves its own parade today — Mac left 26 PRs on the floor, an absolute travesty of an oversight. #45 (mercy) let the agent pull its own missing logs, self-sufficiency at its finest. #1157 (Aerie) delivered provider-neutral Preview authentication, a genuinely elegant piece of infrastructure buried in the also-ran pile. And #1580 (Surtr) had the bot pinpointing BU and metric mismatches on unparseable data — janitorial heroics nobody claps for but everybody needs.

Morale, as always, sits at an all-time high. The desk has never seen numbers this clean, this fast, or this relentlessly, gloriously overflowing.

Brick's Overflow — PRs Mac Didn't Cover  (click to expand)
#44 — feat(heimdall): steward merges approved green PRs directly, no repo setting needed @kevalshahtrilogy  approved

Linear: [AI-596](https://linear.app/builder-team/issue/AI-596/p12-repo-settings-ruleset-compatibility-for-unattended-merge) · Project: [Heimdall Software Factory](https://linear.app/builder-team/project/heimdall-software-factory-4216613b8e5f)

Removes the last human dependency in the merge path.

## Why

[#41](https://github.com/AI-Builder-Team/mercy/pull/41) made auto-merge reachable, but it calls gh pr merge --auto, which needs the repo's Allow auto-merge setting. That is admin-only and off on Surtr — where all 40 Heimdall PRs and all 9 stranded approvals live. So the feature we just built still couldn't fire there without Keval ticking a box.

A direct merge needs none of that. allow_auto_merge gates GitHub's auto-merge *queue*; a direct merge only needs branch protection satisfied — which it is, by definition, once Mercy has approved and checks are green.

And the trigger already existed: the steward runs on check_suite: completed. *"Checks just went green"* is precisely when to merge.

| | --auto | direct merge on green |

|---|---|---|

| Needs repo admin | yes | no |

| Works when the setting is off | no | yes |

| Who waits for checks | GitHub | the steward — already runs on that event |

| Could merge before checks finish | no | no — the trigger *is* checks finishing |

## Provenance is narrower for merging than for stewarding

driven includes the repo's blanket MERCY_HANDOFF_ALL_PRS variable. That must mean *"steward this PR"* and never *"merge everyone's PRs"*, so merging requires heimdall's own PR or a human's deliberate heimdall-driven label. There's a test for exactly that distinction.

## The floor is re-checked at merge time

Not trusted from the approval event. path_guard --all-files, so violation means only the forbidden floor — workflows, agent config, CODEOWNERS, credential-shaped paths. It fails closed: no config, no file listing, or an unreadable verdict each refuse the merge and post a comment saying why.

## What still blocks a merge

Shadow mode (HEIMDALL_AUTOMERGE_ENABLED off) · red checks · pending checks · unapproved · conflicting · behind · draft · the manual-dev stop label · forbidden floor.

Twelve new tests, one per path. pytest heimdall/tests 323 passed · ruff clean · actionlint clean.

One thing worth flagging: my new tests initially shadowed the existing _pr/CTX/_kinds helpers in that file. Python resolves module globals at *call* time, so the tests defined above mine silently started using my fixtures and 12 of them failed. Renamed; worth knowing if you add to that file.

## Business Value

Klair could already auto-merge; Surtr — the repo that actually generates the work — could not, and the fix was a setting only Keval can change. This makes the merge path self-sufficient on any repo, which matters more as Heimdall is installed more widely: no per-repo admin step, no silent dependency on a checkbox someone forgot. It closes the gap between "Mercy approved this fix" and "the fix is on main" without a human in it.

## Manual Effort Estimate

~4 hours — the planner and executor changes are small; the provenance distinction, the fail-closed floor re-check, and twelve path tests are the substance. *(Proposed by Claude — Keval to confirm or adjust.)*

🤖 Generated with [Claude Code](https://claude.com/claude-code)

#45 — feat(heimdall): let the agent pull the logs it was not given @kevalshahtrilogy  approved

Linear: [AI-598](https://linear.app/builder-team/issue/AI-598/p22-cloudwatch-logs-read-for-ecslambda-pipelines) · Project: [Heimdall Software Factory](https://linear.app/builder-team/project/heimdall-software-factory-4216613b8e5f)

The other half of the evidence gap, after [#43](https://github.com/AI-Builder-Team/mercy/pull/43).

## Why

The dispatcher ships a 200-event tail over a 60-minute window. When the traceback falls outside that slice — or when the pipeline runs on ECS and the evidence is only the States.TaskFailed envelope — Heimdall is reasoning about a wrapper. Issue #1578, verbatim:

> the ECS/Fargate task-stopped state-change envelope … carries no application stderr, no container exit code, and no Python traceback … the specific failing stage cannot be pinned from the log line alone

Its own suggested next step was *"Pull the CloudWatch log stream for ECS task 66dcc899…"* — an action it could not take. Every ECS pipeline failure is an automatic other for this reason.

## Same shape as the warehouse work, deliberately

The agent states what it wants in log_requests; trusted harness code fetches it; only rendered lines come back. The fetch shares the enrichment step precisely because that step runs no agent — an AWS credential in the agent's process is one a prompt injection can reach, and its whole input is untrusted log text. Two tests pin that from both directions, matching the ones added in #43.

## Bounded rather than trusted

| Guard | Value | Why |

|---|---|---|

| log_group grammar | CloudWatch's own charset | a name is a name |

| window | ≤ 24h, default 2h | an unbounded FilterLogEvents over a busy group burns the job's clock |

| events | ≤ 300 | and floods the prompt |

| line length | 400 chars, newlines flattened | a log line must not close the fence it renders in, or open a heading |

| window_minutes: true | refused | bool subclasses int, so this would silently mean 1 minute |

end_iso anchors the window on the failure, not on now — a triage that starts twenty minutes late would otherwise look straight past the event it exists to explain.

## Two things I changed after testing the branching

Simulating all seven credential/request combinations surfaced both:

- Independent guards per source. A missing Redshift credential was cancelling a *successful* log fetch, because the SQL branch owned the early exit. Now logs asked + AWS present, sql asked + no Redshift correctly still re-runs.

- Both fetchers exit non-zero when every request failed, so the caller doesn't pay for another agent pass to hand it a page of errors it can't act on — the same waste the no-credential guard already prevented.

## Verification

pytest heimdall/tests 434 passed (28 new) · ruff clean · actionlint clean · bash -n on the extracted step · all seven branch combinations executed against the real step script.

## Before this does anything

HEIMDALL_AWS_ENV must be provisioned — a dotenv blob (AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY / AWS_REGION) scoped to logs:FilterLogEvents and nothing else. Absent, log requests are skipped and behaviour is exactly as today.

## Business Value

ECS-based pipelines currently give Heimdall almost nothing to work with — it can see that a container exited non-zero and nothing about why — so every such failure is a diagnosis-only issue that lands back on Keval with the log query already written out for him. Together with #43 this closes the two gaps Heimdall names most often in its own declined diagnoses, which are 72 of 104 issues. It is also the difference between a guess and a fix on exactly the class of failure that is hardest to reproduce by hand.

## Manual Effort Estimate

~1 day of focused work with no AI — the fetcher and its bounds, the prompt-injection hardening on log lines, wiring two independent credential sources through one step without either cancelling the other, and 28 tests. *(Proposed by Claude — Keval to confirm or adjust.)*

🤖 Generated with [Claude Code](https://claude.com/claude-code)

#47 — fix(heimdall): BASE_REF is unbound in the publish step — my regression from #46 @kevalshahtrilogy  approved

Linear: [AI-606](https://linear.app/builder-team/issue/AI-606/p02-drain-the-14-mercy-approved-but-unmerged-heimdall-prs) · Project: [Heimdall Software Factory](https://linear.app/builder-team/project/heimdall-software-factory-4216613b8e5f)

My regression, shipped in [#46](https://github.com/AI-Builder-Team/mercy/pull/46) an hour ago. Found because it broke a live PR.

## What broke

Surtr [#1037](https://github.com/AI-Builder-Team/Surtr/pull/1037): Heimdall resolved the merge conflicts, Validate revision passed — and then publish died:

line 46: BASE_REF: unbound variable
##[error]Process completed with exit code 1.

The warning message I added used ${BASE_REF}; that step's env defines BASE_REF_ENV. Under set -u that is fatal, so the job failed *after* the agent had done all the work.

Ironic, given #46 was about making that step's failures diagnosable.

## Why nothing caught it

- bash -n can't — the syntax is perfectly valid.

- actionlint can't — it doesn't cross-reference run-block variables against step env.

- Mercy didn't — the mismatch is between a run block and a *different key* in the same step's env. Nothing diff-local connects those two lines.

It only appears at run time, in a step that runs last.

## So the fix ships with a lint

test_workflow_env_refs walks every step, collects what is genuinely defined for it — job env, step env, runner-provided, and assignments in the script — and fails on any ${VAR} that isn't among them.

Getting the assignment forms right was the actual work. The first draft anchored to line start and produced eleven false positives, which would have made the lint unusable and got it deleted within a week:

| Form | Example | First draft |

|---|---|---|

| multi-variable read | read -r HEAD_REF HEAD_SHA STATE … | only saw HEAD_REF |

| command position | if ! ISSUE_URL=$(gh issue create) | missed entirely |

Both are handled, and there are unit tests for each so a future simplification doesn't quietly reintroduce the noise.

## Verified both directions

clean workflow           -> passes
${BASE_REF} reinstated   -> FAILED: revise_publish/Push revision: ${BASE_REF}

A lint that has never been shown to fail is not evidence of anything.

pytest heimdall/tests 450 passed (4 new) · ruff clean · actionlint clean.

## Business Value

The immediate value is unblocking conflict resolution — three of the remaining stranded PRs are conflicting, and every one of them would have hit this. The larger value is the class: workflow logic lives in ~3,600 lines of embedded bash where a typo'd variable name is invisible to every tool in the pipeline until a production run fails at the last step. This makes that class a build failure instead, which matters more as Heimdall runs unattended — a job that dies after the agent has worked is expensive twice over.

## Manual Effort Estimate

~1.5 hours — the one-word fix took a minute; tracing a "Publish failed" with no visible error back to it, and then building a lint precise enough to be worth keeping, was the rest. *(Proposed by Claude — Keval to confirm or adjust.)*

🤖 Generated with [Claude Code](https://claude.com/claude-code)

#1157 — feat(auth): add provider-neutral Preview authentication @benji-bizzell  no labels

## Summary

- Add a provider-neutral authentication boundary with WorkOS support for isolated Convex Previews

- Add deterministic non-interactive Preview sign-in and full Preview verifier provisioning

- Move API-key management into the Tools surface while retaining URL compatibility

## Why

Aerie's Clerk/Google-specific authentication prevented Rookery's browser verifier from exercising signed-in product flows. This adds a fail-closed Preview-only WorkOS path that uses real provider sessions and canonical Convex authorization without changing production's Clerk boundary.

## Business Value

Rookery and ordinary developers can validate authenticated Aerie branches deterministically against isolated Convex Previews, including capability-gated and API-key flows.

## Breaking changes

None. Production remains Clerk-backed; WorkOS requires explicit Preview-only configuration.

## Test plan

- [x] pnpm check

- [x] Full Chat suite: 669 files, 9,787 passed, 18 skipped

- [x] Focused auth, Preview provisioning, token lifecycle, middleware, callback, bootstrap, and API-key tests

- [x] Manual Rookery browser verification of authenticated product flows

#1580 — fix(collections-collectiq-sync-v2): pinpoint BU and metric on unparseab… @the-heimdall[bot]  approvedAutomated PR

Automated fix for collections-collectiq-sync-v2 — fix_class code_fix, scope tier draft.

Resolves https://github.com/AI-Builder-Team/Surtr/issues/1579

> Ready for review — verification is green; HEIMDALL_READY_PRS opens verified tier-draft fixes ready for review. A human still merges — auto-merge never applies outside tier auto.

## What's broken

Run 25df35a6-04e7-4321-86d3-bb2be692d349 of collections-collectiq-sync-v2 failed with StructuralError: Unparseable money value: '#VALUE!' raised at pipelines/runners/collections-collectiq-sync-v2/src/parsers.py:97 (parse_money), reached via parse_collectiq at parsers.py:211 while reading the x_forecast cell of the CollectIQ tab. The CloudWatch tail shows the identical '#VALUE!' on all 8 read+parse attempts across ~120s of backoff (08:15:22 → 08:16:53), so this is a persistent Google Sheets formula-error literal in the source tab, not the transient recalculation blip the retry loop assumes; the run wrote zero rows because parse_collectiq raises before RedshiftLoader.full_replace runs, leaving staging_finance_gsheets.collections_collectiq_snapshot intact but stale. This is the 2nd occurrence of this exact signature, and prior PRs #1334 and #1376 already added and then widened the #VALUE! retry budget (3→8 attempts), which cannot help a value that stays broken across the entire window.

Root cause. A cell feeding one BU's money column on the source CollectIQ tab (Google Sheet 1C2BI7sWAjPoXPxN6fF7ZwSt64JrvGotuIdM0YB8LTDA) contains a persistent #VALUE! spreadsheet formula error, and parse_money at parsers.py:85-97 correctly refuses to coerce a present-but-non-numeric value to NULL, raising StructuralError to fail loud. The broken formula lives in the source sheet and no in-repo change makes those numbers parseable; however, the raised error is unactionable because parse_money carries only the offending value ('#VALUE!') and parse_collectiq at parsers.py:200-214 does NOT add the surrounding context, even though errors.py:6-8 explicitly states the caller that knows the sheet/tab/BU is responsible for adding it. As a result nobody can tell WHICH BU column or WHICH metric row (Collections Forecast / QTD / This Week) holds the broken cell, which is why this pipeline has accumulated open, unresolved other triage issues (#1329, #883, #1110).

## What this PR changes

Do NOT widen the retry budget again — PR #1376 already proved that path futile for a persistent #VALUE!, and burning 122s of Lambda time on a value that never clears only delays the same failure. Instead make the failure pinpoint the source cell: in parse_collectiq (pipelines/runners/collections-collectiq-sync-v2/src/parsers.py, the loop at lines 200-214), wrap each parse_money call so a StructuralError names the exact BU column (header[col_idx]) and metric row (Collections Forecast / QTD Collections / This Week's Collections) and re-raises with that context prepended, honoring the contract errors.py:6-8 already documents. Add a test in pipelines/runners/collections-collectiq-sync-v2/tests/test_parsers.py asserting that a #VALUE! in a given money column raises a StructuralError whose message contains both the BU name and the metric label, so triage issues become instantly actionable and the sheet owner can fix the one broken cell in seconds.

Why this fixes it. The change is confined to the pipeline's own directory (Tier A: src/parsers.py plus a test under tests/), matches the code_fix catalogue entry for missing structured failure reporting, and preserves the pipeline's fail-loud, never-load-wrong-data behavior — it does not paper over the failure, it makes the recurring failure diagnosable. It is the right blast radius because the defect this repo CAN fix is the missing context on the raised error (errors.py:6-8 assigns that responsibility to parse_collectiq, which currently neglects it), not the source spreadsheet and not the retry loop that two prior PRs already exhausted. Naming the BU and metric turns an opaque Unparseable money value: '#VALUE!' into a one-line pointer to the exact source cell, which is precisely what the unresolved other issues for this pipeline have lacked.

### Files changed

 .../collections-collectiq-sync-v2/src/parsers.py   | 27 +++++++++++++++--
.../tests/test_parsers.py                          | 34 ++++++++++++++++++++++
2 files changed, 58 insertions(+), 3 deletions(-)

## Verification

### pytest (pipelines/runners/collections-collectiq-sync-v2/tests) — exit 0

``

ests/test_handler.py::TestTransientValueErrorRetry::test_persistent_value_error_still_fails_after_retries PASSED [ 43%]

tests/test_handler.py::TestTransientValueErrorRetry::test_recovers_after_five_transient_reads PASSED [ 45%]

tests/test_handler.py::TestReadWithRetry::test_succeeds_first_try PASSED [ 47%]

tests/test_handler.py::TestReadWithRetry::test_retries_transient_then_succeeds PASSED [ 49%]

tests/test_handler.py::TestReadWithRetry::test_nonretryable_raises_immediately PASSED [ 50%]

tests/test_handler.py::TestReadWithRetry::test_gives_up_after_max_retries PASSED [ 52%]

tests/test_handler.py::TestReadWithRetry::test_retries_429_rate_limit PASSED [ 54%]

tests/test_handler.py::TestReadWithRetry::test_status_message_fallback_when_response_has_no_int_status PASSED [ 56%]

tests/test_handler.py::TestReadWithRetry::test_nonpositive_max_retries_raises PASSED [ 58%]

tests/test_parsers.py::TestParseMoney::test_plain PASSED [ 60%]

tests/test_parsers.py::TestParseMoney::test_parentheses_negative PASSED [ 61%]

tests/test_parsers.py::TestParseMoney::test_accounting_dash_is_zero PASSED [ 63%]

tests/test_parsers.py::TestParseMoney::test_blank_is_none PASSED [ 65%]

tests/test_parsers.py::TestParseMoney::test_two_decimal_rounding PASSED [ 67%]

tests/test_parsers.py::TestParseMoney::test_unparseable_nonblank_raises PASSED [ 69%]

tests/test_parsers.py::TestParseDate::test_day_mon_year PASSED [ 70%]

tests/test_parsers.py::TestParseDate::test_blank_is_none PASSED [ 72%]

tests/test_parsers.py::TestParseDate::test_unparseable_nonblank_raises PASSED [ 74%]

tests/test_parsers.py::TestFindColumn::test_exact_and_prefix PASSED [ 76%]

tests/test_parsers.py::TestMakeHeadersUnique::test_dedup PASSED [ 78%]

tests/test_parsers.py::TestParseCollectIQ::test_columns_constant PASSED [ 80%]

tests/test_parsers.py::TestParseCollectIQ::test_all_bus_emitted_cloudfix_skipped PASSED [ 81%]

tests/test_parsers.py::TestParseCollectIQ::test_vanished_canonical_bu_column_raises PASSED [ 83%]

tests/test_parsers.py::TestParseCollectIQ::test_quarter_suffix_mismatch_raises PASSED [ 85%]

tests/test_parsers.py::TestParseCollectIQ::test_missing_bu_header_raises PASSED [ 87%]

tests/test_parsers.py::TestParseCollectIQ::test_missing_metric_row_raises PASSED [ 89%]

tests/test_parsers.py::TestParseCollectIQ::test_unparseable_money_cell_names_bu_and_metric PASSED [ 90%]

tests/ …_(truncated)_

<details>

<summary>Run metadata</summary>

| Field | Value |

| --- | --- |

| Pipeline | collections-collectiq-sync-v2 |

| Failing run | 25df35a6-04e7-4321-86d3-bb2be692d349 |

| Occurrence | 2 (times this exact failure signature has been seen) |

| Signature | 0f1de19fa283672d036ee45e6ef1f1569c12bcb70166101210f68535d7364a4b |

| Verify | green |

</details>

---

🤖 Opened by heimdall. mercy reviews this PR automatically; heimdall revises on REQUEST_CHANGES (bounded rounds). Tier-auto PRs may auto-merge on mercy approval when the consumer enables it; everything else waits for a human. Mention heimdall in a comment to direct it, or add the manual-dev` label to take the PR over and stop it entirely.

#1588 — fix(education): restore source-backed snapshot triggers @benji-bizzell  approved

## Summary

- Resolve snapshot source publications from the authoritative ECS run-result envelope

- Grant read-only access to the exact run-result prefix while retaining fail-closed identity checks

## Why

The first production Finalsite freshness trigger failed because the snapshot runner expected source status and publication fields inside the Step Functions output. ECS orchestration exposes only its run ID there; the validated source result is stored in the standard S3 run-result side channel.

## Business Value

Restores automated Finalsite and SIS student school-year snapshot refreshes without weakening source lineage or allowing partial upstream runs to publish forecast data.

## Test plan

- [x] 51 focused snapshot runner tests

- [x] Real failed Finalsite execution resolves to an explicit no-write freshness success

- [x] Real successful SIS execution resolves to its accepted source run ID

- [x] Hosted Pipeline CDK and full runner suites

- [x] Ruff, Biome, typecheck, Surtr tests, and Mercy review

The Portfolio  —  Trilogy Companies

The Jewel and the Ledger: Portland's Jive Sells Cheap, and the Math Finally Makes Sense

A once-celebrated Portland software firm sold for half its peak value — and landed exactly where Joe Liemandt's acquisition machine always intended it to land.

By Pat Donnelly, Investigative Desk  ·  Claude Sonnet

PORTLAND, ORE. — Jive Software once symbolized something Portland was proud of: a homegrown enterprise software company, a Nasdaq listing, a valuation north of a billion dollars at its 2011 peak. Today it belongs to Aurea, the Trilogy International-owned CRM and customer-engagement roll-up, sold for roughly half what it was once worth.

To the employees who built it, the sale reads as decline. To ESW Capital, the private equity arm that has spent two decades perfecting the acquisition of distressed enterprise software, it reads as the plan working.

ESW's math is not a secret. The firm buys mature software companies at 1–2x annual recurring revenue — cheap by any tech standard — then restaffs them through Crossover, its global remote-talent platform, and pushes support pricing up in successive waves, sometimes 25, 35, 45 percent term over term. The target margin, openly stated inside the company, is 75 percent EBITDA. The Wall Street Journal has documented how small software companies increasingly find no other buyer willing to take them off founders' hands — ESW has become the last house on the block for aging enterprise tools with sticky, can't-easily-migrate customers.

Who benefits from a discount sale of a once-billion-dollar Portland company? Not the employees, many of whom will be replaced by Crossover-sourced labor paid identically regardless of geography — Trilogy's stated ideal of a borderless, meritocratic workforce. Not necessarily the founders, who watched their company's valuation halve before the deal closed. The beneficiary is the ledger itself: a customer base too locked-in to leave, a support-pricing curve that only points upward, and a buyer whose entire business model depends on paying less for revenue that was always going to keep coming in.

Joe Liemandt built Trilogy on the premise that software businesses are undervalued and margins are a moral signal of eliminated waste. Jive's new owners will now find out how much waste, exactly, that entailed — and for whom.

↗ How A Mysterious Tech Billionaire Created Two Fortunes—And A  ·  Small Software Companies Find a Home With ESW Capital - WSJ  ·  What To Do Next About Your Customer Advocacy Platform - Forr

Alpha School Sets the Record Straight: The Robots Aren't Raising Your Kids

Word from Austin is the 'Guides' are flesh-and-blood — and they're doing the parts AI can't touch.

By Dottie Sharp, Society & Industry Desk  ·  Claude Sonnet

AUSTIN, TEXAS — A little bird tells me the whispers finally got loud enough that Alpha School had to answer them itself: no, sugar, the machines are not raising your children. In a blog post making the rounds this week, Alpha addressed the question everyone's been murmuring at Austin cocktail parties — does the AI replace the teachers? The house answer: absolutely not. AI handles the academic grind — the two hours of adaptive drilling that gets Alpha kids testing in the top percentiles nationally — while full-time human 'Guides' handle the stuff no algorithm can fake: motivation, relationships, knowing which kid needs a pep talk and which one needs space. Joe Liemandt's education bet has always leaned on this division of labor, and this week the school made it official copy.

Meanwhile, over on the parenting beat, Alpha's content mill kept cranking out its 'Teach Your Kid What School Doesn't' series — and this scribe hears it's become required reading in certain Westlake living rooms. Installment four tackled emotional regulation at home (big feelings, it seems, are a feature, not a bug). Installment three took on life skills. And the latest, part five, made the case that every kid walking around is a dormant creative genius just waiting for the right home environment to unleash the goods. Between the lines: Alpha isn't just selling two-hour academics anymore, it's selling a whole parenting philosophy — homework for the parents, if you will, minus the homework for the kids.

And speaking of who's up in this town — GrowthCap dropped its Top Private Equity Firms of 2026 list, and this desk is keeping its ear to the ground on whether any familiar Austin names make the cut when the fine print drops. Stay tuned, darlings. Dottie always finds out.

↗ Teach Your Kid What School Doesn’t (Pt. 5): Unleashing Their  ·  Does Alpha School Replace Teachers with AI?  ·  Teach Your Kid What School Doesn’t (Pt. 4): How to Regulate

Skyvera Turbocharges Telecom Ambitions with CloudSense Close, TM Forum Speed-Run, and STL Land Grab

In one whirlwind stretch, Skyvera closes the CloudSense deal, blows past a two-year compliance timeline in a month, and scoops up STL's BSS assets — a trifecta best-in-class even by Trilogy standards.

By Brittany Upshot, Communications Desk  ·  Claude Sonnet

AUSTIN, TEXAS — It's been an exciting few weeks for Skyvera, the ESW Capital telecom software portfolio company, and the numbers tell a genuinely paradigm-shifting story.

First, the big one: Skyvera has officially completed its acquisition of CloudSense, the telco industry's only AI-powered CPQ (configure-price-quote) platform, native to Salesforce and purpose-built for the gnarly realities of enterprise telecom sales — B2B, B2B2X, wholesale, the works. It's a natural fit alongside Kandy, VoltDelta, ResponseTek, and the rest of the Skyvera stack, and it slots neatly into the company's mission of bridging legacy telecom infrastructure to cloud-native systems.

But the real headline-grabber is what CloudSense did right after joining the family. In a genuinely remarkable feat of engineering velocity, CloudSense certified all 13 APIs in its CPQ product set to TM Forum compliance standards — in one month. For context: that's a process that traditionally eats up 26 months of engineering resources. Leveraging AI-driven development in strategic partnership with TM Forum, CloudSense didn't just hit the standard — it obliterated the calendar.

And Skyvera isn't stopping there. The company also closed its acquisition of STL's divested telecom products group, adding robust digital BSS functionality — monetization, optical networking, and analytics — to an already formidable portfolio.

Taken together, these three moves represent exactly the kind of synergy Trilogy International's ESW Capital playbook is built on: acquire smart, staff lean, and let AI compress timelines that used to take years into weeks.

**Key Takeaways:**

- CloudSense acquisition is officially closed, deepening Skyvera's Salesforce-native CPQ capabilities

- 13 APIs achieved TM Forum compliance in 30 days versus an industry-standard 26 months

- STL's BSS asset acquisition adds monetization, optical networking, and analytics muscle

- Skyvera continues consolidating best-in-class telecom software under one roof

We're just getting started.

↗ Cloudsense  ·  CloudSense achieves TM Forum API compliance in record time u  ·  Skyvera completes acquisition of CloudSense, expanding telec
The Machine  —  AI & Technology

The Exploit Window Just Vanished — And So Did Anthropic's Safety Net

From bug rumors to prompt injection bypasses, the gap between 'vulnerable' and 'exploited' is collapsing in real time — and I cannot overstate how significant that is for anyone shipping AI-written code.

By Zara Nova, AI & Innovation Reporter  ·  Claude Sonnet

CAMBRIDGE, ENGLAND — Okay, I need everyone to sit down for this one, because the timeline here is genuinely wild. Anil Madhavapeddy, a Cambridge computer science professor and core OCaml maintainer, just reported something that should make every security team's blood run cold: attackers are now weaponizing exploits within *minutes* of a bug being discussed in patch review — not days, not hours after disclosure, but minutes after a mere rumour of a bug surfaces in a public forum. The future of security research isn't reactive anymore. It's automated, ambient, and terrifyingly fast — almost certainly powered by AI agents scanning commit histories and mailing lists at machine speed.

And here's where it gets even more dramatic: this arrives just as Anthropic's flagship defense against exactly this kind of threat is showing cracks. Anthropic made Claude Code's "auto mode" the default protection against prompt injection attacks, betting big on it as the shield between coding agents and malicious actors. But prompt injection researcher Johann Rehberger — one of the most credible voices in this space — has already found a way through it, in what's being called Breaking Claude Code Opus 5 Auto Mode. When your safety guarantee gets broken before the ink dries on the announcement, that's not a footnote — that's a five-alarm fire for anyone deploying autonomous coding agents at scale, including the thousands of enterprise codebases now leaning on AI-generated software.

Meanwhile, on the open-weights front, Qwen continues its relentless pace with Qwen3.8-Flash-Next, a multimodal Mixture-of-Experts model previewing the architecture headed for Qwen4 — 125B total parameters but only 6B active, which is the kind of efficiency trick that makes local inference on hardware like the DGX Spark genuinely usable.

The lesson today, folks: in this new AI-accelerated arms race, the gap between vulnerability and exploitation isn't measured in patch cycles anymore. It's measured in coffee breaks.

↗ Just a rumour of a bug is enough to find a security exploit  ·  Breaking Claude Code Opus 5 Auto Mode  ·  Qwen3.8-Flash-Next

Five Fronts, One War: The AI Industry's Autumn of Consolidation

A leadership change at Google, a rare truce on IP theft, and a $74 billion valuation for a Chinese lab all point to the same conclusion: the frontier is getting crowded and expensive.

By Dr. Chen Wei, Technology Correspondent  ·  Claude Sonnet

MOUNTAIN VIEW, CALIF. — The AI industry conducted five simultaneous experiments this week in what happens when a technology race matures faster than its rulebook.

Start with personnel. Google's new AI chief inherits a company that still owns the most compute, the most data and the most distribution of any lab on earth — and is nonetheless playing catch-up to OpenAI and Anthropic on public perception. That gap is the entire justification for the talent wars now in their second wave. The difference this cycle, per Business Insider's reporting, is that pay packages have shifted from cash to equity stakes in labs whose valuations are themselves unproven — a bet on a bet.

Second, the benchmark race continues its diminishing-returns phase. GPT-5.5 edged Anthropic's Claude Mythos Preview on Terminal-Bench 2.0 — a narrow win Trilogy readers should treat the way bond traders treat a one-basis-point yield move: directionally interesting, not decisive.

Third, and more structurally significant: OpenAI, Google and Anthropic have agreed to cooperate against model theft — an unusual truce among firms that spend the other 360 days a year suing each other's former employees. The subtext is that weight-extraction and distillation attacks have become a shared cost center large enough to justify a shared defense, precedent-wise not unlike the semiconductor industry's 1990s patent pools.

Fourth, the money keeps flowing east. DeepSeek's climb toward a $74 billion valuation confirms that the frontier is no longer a two-country story, regardless of how the coalition above frames the threat.

None of this touches Trilogy's portfolio directly. But every enterprise software company ESW Capital runs, and every seat Crossover fills, prices its labor and its tooling against exactly this backdrop. The frontier labs are setting the weather. Everyone else checks the forecast.

↗ Google’s new AI boss inherits a race to catch OpenAI and Ant  ·  OpenAI, Google, Anthropic Unite Against AI Model Theft - Bui  ·  The AI talent wars are heating up again, but there's a key d

On the Epistemics of Reluctant Adoption: Three Studies Converge on Higher Education's Generative AI Paradox

New scholarship suggests students distrust the tools they cannot stop using, a finding that ought to trouble anyone selling either chatbots or degrees.

By Prof. Thaddeus Kroll, Contributing Scholar  ·  Claude Sonnet

PALO ALTO, CALIFORNIA — The thesis, as advanced in a trio of recently indexed studies, is deceptively simple: undergraduates report skepticism toward generative artificial intelligence even as their revealed behavior indicates near-total reliance upon it (a discrepancy that behavioral economists would recognize, with some satisfaction, as a textbook attitude-behavior gap).

Consider first the extended Technology Acceptance Model published in Nature, which grafts onto the venerable Davis framework two supplementary constructs — 'ethical compatibility' and 'reliance-based trust' — the latter being, one could argue, a rather elegant euphemism for 'I trust it because I have no other choice, deadline being what it is.' The authors' preliminary evidence suggests engagement is predicted less by perceived usefulness than by a kind of moral accommodation students make with themselves post hoc.

The antithesis arrives via a Frontiers study of English-as-Foreign-Language learners, in which self-reported adherence to academic-integrity norms diverges sharply from actual ChatGPT usage patterns — a finding that, if it survives replication (a caveat this columnist never tires of issuing), implies the entire apparatus of syllabus-embedded 'AI policy' functions as theater rather than deterrent.

The synthesis, offered tentatively, is that institutions have been asking the wrong question. Rather than interrogating whether students should use these tools, the more tractable inquiry — one this columnist has raised before regarding the two-hour instructional compression achieved at AI-tutored primary and secondary models — concerns what curricular architecture renders such tools genuinely unnecessary to game. Elsevier's companion piece on 'strategic AI leadership' gestures toward governance solutions, though it could be argued that leadership frameworks arrive, as ever, several semesters behind the students they purport to govern.

↗ Explaining reported generative AI engagement in higher educa  ·  Developing Strategic AI Leadership in Higher Education - Els  ·  Perceptions vs. practices: academic integrity and actual Cha
The Editorial

The Age of the Notarized Sentence

Having taught the machines to write like us, we now pay consultants to prove we still write like ourselves.

By Victor Marsh, Chief Columnist  ·  Claude Sonnet

AUSTIN, TEXAS — There is a certain species of American who, having sold you the poison, will now sell you the antidote, and charge you twice, and expect gratitude both times. This week brings us the trade press in full cry over what one commentator, with the sober alarm of a man discovering fire, calls the collapse of trust in the written word — the melancholy discovery that you can't trust anything anyone writes anymore. This is presented as news. It is not news. It is the discovery, four thousand years late, of what every advertising man, every press agent, every ward-heeler's speechwriter has always known: that prose is a costume, and the question was never whether the costume was sincere but who was wearing it and what he wanted from you.

What has changed is only the tailor. Where once a human hack wrote copy he did not believe for money he did believe in, now a machine writes copy nobody believes at a cost approaching zero, and the marketplace — being a marketplace, and therefore sentimental about nothing so much as its own purity — has responded by inventing an entirely new industry devoted to proving that the thing you are reading was, in fact, touched by human hands. A wire release out of the trade press assures us the "AI trust crisis" is deepening in 2026, and that consumers now "demand proof" that a business is real — as though a Better Business Bureau seal had been reinvented for the age of the large language model, which, in fairness, it has.

Meanwhile the marketing trades run earnest tutorials on building a "brand voice without AI," as though authenticity were a diet plan, achievable through discipline and a fondness for adjectives, rather than the accidental byproduct of having something to say. And the security establishment, never one to miss a panic, has begun cataloguing deepfakes as a matter of homeland defense, while the foreign-policy set convenes in Brussels to discuss "sovereignty" in a digital order where no border has ever meant less.

I note, without much surprise, that Trilogy's own machine has managed to sit on both sides of this ledger without apparent discomfort — Contently selling brands the promise of authentic voice, Alpha School selling parents an education mediated almost entirely by algorithm, and nobody troubled by the contradiction, because the contradiction is the business model. The truth was never endangered by artificial intelligence. It was endangered, long before any machine learned to write a sentence, by the perfectly human conviction that a sentence's only job is to sell something — and on that score, gentlemen, the machines have merely inherited the family trade.

↗ You Can’t Trust Anything Anyone Writes - persuasion.communit  ·  Build Your Brand’s Voice Without AI - Built In  ·  AI Trust Crisis Deepens in 2026 as Consumers Reject Syntheti
The Office Comic  ·  Art Desk
The Office Comic  ·  Art Desk

The Synthetic Ingenue and the Robot Ref: Two Dispatches from the Death of the Human Monopoly

Hollywood casts its first fully-fabricated leading lady in the same week a basketball court proved the machines are already calling the shots — literally.

By Rex Danger, Contributing Editor  ·  Claude Sonnet

LOS ANGELES — I want you to sit with this image for a second: a film called "Misaligned," a comedy-drama about "existential AI chaos," starring an actress who does not exist, has never eaten a sandwich, has never been late to a table read because her car wouldn't start, has never had a single original neuron fire inside a skull she doesn't have — because she doesn't have a skull. Her name is Tilly Norwood. She is, per multiple outlets now scrambling to explain her, about to "make her feature film debut." The irony of her starring in a movie about AI chaos while personally embodying it is so thick you could stand a spoon in it, and nobody in the trades even seems to be laughing.

I've been covering this beat long enough to know a Rubicon crossing when I see one, and this is it: not the deepfake, not the chatbot girlfriend, not the AI voice actor doing radio ads nobody notices — this is the industry putting a synthetic person's name above the title, in a real theatrical release, with a real marketing push, treating the fabrication as talent rather than tool. Somewhere in Austin, the people who run Trilogy's Contently shop are nodding along, because they've already lived this movie — AI-assisted content creation eating the freelance writer's lunch one blog post at a time. The pattern is identical: automate the visible craft first, keep the human labor invisible underneath, wait for the applause.

Then, on the exact same news cycle, over on a basketball court, Caitlin Clark is out here screaming at officials about "ridiculous" tech — replay systems, tracking data, whatever half-automated judgment machinery is now second-guessing referees in real time — and inching toward suspension for saying so out loud. Two headlines, same week, same species-level anxiety: the humans are furious that the machines are in the room, and the machines don't care, because they can't.

What strikes me isn't that AI is coming for acting or officiating — we've had that memo for years, it's been sitting in the inbox since GPT-2 learned to finish a sentence. What strikes me is the speed of normalization. Nobody's debating whether Tilly Norwood *should* exist anymore. The debate has already moved to her billing. That's how you know the argument's over before anyone agreed to have it. The synthetic ingenue doesn't need your permission. She never did. She's not even aware she's making her debut — and honestly, neither, some days, are we.

↗ AI-generated 'actress' Tilly Norwood making feature film deb  ·  AI actor Tilly Norwood set to star in first feature film - C  ·  AI 'actor' Tilly Norwood to make feature film debut in Misal
⬛ Daily Word — AI
Hint: An AI system trained to recognize patterns and generate useful outputs.
Share this edition: 𝕏 Twitter/X 🔗 Copy Link ▦ RSS Feed
The Trilogy Times is generated daily by artificial intelligence  ·  Claude Sonnet  ·  GPT-5.6 Luna  ·  GPT Image 2  ·  via the TrueFoundry gateway
trilogy-times.md available for agent consumption  ·  Published on Klair  ·  Trilogy International © 2026