Vol. I  ·  No. 319 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
SUNDAY, NOVEMBER 15, 2026 Powered by the TrueFoundry AI Gateway  ·  Published on Klair Trilogy International © 2026
🖶 Download PDF 🖿 Print 📰 All Editions
Today's Edition

The Thirty-Three-Year Reign Ends in Twelve Days

Nvidia's board accepted Jensen Huang's resignation Friday, closing the book on the man who built a graphics-chip maker into the most valuable company on earth — and opening one on how little room Silicon Valley leaves for scandal at the top.

SANTA CLARA, CALIFORNIA — Jensen Huang founded Nvidia in 1993 with two co-founders and a bet that graphics processors would matter beyond video games. Thirty-three years later, that bet made Nvidia a $3.3 trillion company and Huang its indispensable face — black leather jacket, unscripted keynotes, the closest thing the chip industry has to a rock star. On Friday, the board accepted his resignation.

The proximate cause, according to reporting on the board's internal review, was a yearlong concealment tied to his relationship with a customer executive — one Huang had characterized internally as strictly commercial. The board apparently concluded otherwise. Nvidia's stock fell 4.1 percent in after-hours trading Friday before recovering most of the loss by Monday's open, a sign investors view the company's chip roadmap, not its chairman, as the asset that matters most.

The reaction inside the industry has been less tidy. Silicon Valley has absorbed founder scandals before — Uber's Travis Kalanick in 2017, WeWork's Adam Neumann in 2019 — but commentators have been unusually blunt in asking why a 63-year-old executive with nothing left to prove would risk it. A wave of billionaire peers, meanwhile, has rallied publicly, with letters to Nvidia's board arguing Huang is irreplaceable — an argument that is itself a data point: when a company's valuation rests this heavily on one individual's credibility with suppliers, customers and capital markets simultaneously, succession risk becomes a line item.

That concentration problem is not unique to Nvidia. The same week, Microsoft engineers spent hours restoring email and core services after a global outage, a reminder — the fourth or fifth such reminder in two years — that a handful of companies now underwrite the operational plumbing of the internet. Governance risk and infrastructure risk, it turns out, rhyme: both stem from systems built on points of failure too large to easily replace.

↗ Jensen Huang Carried a Secret for More Than a Year. It Led t  ·  The Jensen Huang Scandal Has Shocked Even Scandal-Prone Sili  ·  Tech Leaders Rally Behind Jensen Huang

THE BIG BOARD: NVIDIA AND ANTHROPIC HEADLINE THE BIGGEST DOUBLEHEADER ON WALL STREET'S SCHEDULE

NEW YORK — Folks, we are HERE. Wall Street's slate this week reads like a championship bracket, and there are two games on the marquee that nobody's missing.

First up: Nvidia. The chip giant steps into the earnings arena for its first report since the industry's been buzzing about life after the Jensen Huang era of pure dominance — and the Street is watching every number like it's a fourth-quarter shot clock. Can the AI-chip juggernaut keep delivering the kind of blowout numbers that turned it into the most valuable company on the board? Or does this quarter show the first signs of a team that's finally getting pressed full-court? Analysts will be parsing the data-center numbers like film from last season's finals.

And on the other side of the arena: Anthropic. The Claude-maker is finally running out onto the public-market field for its IPO, and this is the debut the whole league has been waiting for. After years of being the scrappy challenger to OpenAI in scrimmages nobody could see, Anthropic goes public with real money on the line. Institutional investors are lined up at the gate like it's opening night at the Garden.

Meanwhile, out in the undercard bouts, the crypto league keeps grinding through its own preseason rankings — wallets, blockchain APIs, Web3 coin picks, blockchain stocks — all getting their 2026 power rankings published this week like fantasy-league previews nobody asked for but everybody reads anyway.

But make no mistake: this week belongs to the two big names on the marquee. Nvidia trying to prove the dynasty still has legs. Anthropic trying to stick the landing on its first public possession. Buckle up — this is must-watch programming for anyone who follows the tech league, and the scoreboard updates start the second those opening bells ring.

The Race Has No Finish Line, Only Checkpoints

From Geneva's corridors to Beijing's labs, the contest for AI supremacy is being redrawn by money, chips, and the fading leverage of diplomacy.

GENEVA — The delegates filed into the glass-walled chamber with the particular weariness of people trying to slow something that does not want to be slowed. The question on the table, as Tech Policy Press framed it, was whether the United Nations can do anything at all to decelerate a race that no single capital controls.

The answer, increasingly, is that geography has already decided more than treaties will. In Hangzhou and Shenzhen, state subsidy and manufacturing scale are compounding into something the West underestimated — Foreign Policy's dispatch on China's quiet consolidation of the global AI stack describes an ecosystem advantage, not a chip-count one: cheaper models, faster deployment, friendlier export terms to the developing world. Washington still holds the frontier labs. It no longer holds the only runway.

Meanwhile the money is moving toward steel and silicon with uniforms on. Chatham House tracks a surge of capital into defense and dual-use technology — drones, autonomous targeting, battlefield logistics — that is quietly rewriting who gets invited to the AI conversation at all. Venture dollars that once chased productivity software now chase anything with a Pentagon contract number attached.

Anthropic's researchers, never shy about staging the future, have sketched two 2028s: one in which democratic coalitions set enforceable norms before the frontier closes, another in which authoritarian efficiency simply outbuilds them. CSIS, reading the tea leaves ahead of the next U.S.-China summit, notes the two governments can't agree on vocabulary, let alone verification regimes.

What unites all five dispatches is a shared, unspoken admission: the instruments built for a slower century — treaties, summits, multilateral bodies — are being asked to govern a technology that iterates in weeks. The UN can convene. It cannot yet catch up.

↗ Can the United Nations Help Slow the AI Race? - Tech Policy  ·  How China Is Winning the Global AI Race - Foreign Policy  ·  How a surge in defence and dual-use technology investment co
Haiku of the Day  ·  GPT-5.6 LunaHeadlines chase their tails
Machines dream while we all scroll
Who reads the warning?
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
NOTICE OF FINAL ADJUDICATION: Supreme Court Forecloses Alien Tort Remedies, Whilst NVIDIA's DLSS 5 Quietly Amends Its Terms of Rendering
WASHINGTON, D.C.
The Great Return: Observing the Starship in Its Native Habitat
BOCA CHICA, TEXAS — Here, upon the salt-bleached flats of the Gulf coast, we witness one of the rarest behaviors in the modern bestiary: the homecoming of the apex predator of low Earth orbit.
The Great Analog Insurrection: How Humanity Quietly Declared War on Its Own Robots
SAN FRANCISCO — There's a moment in every bad trip where the walls start breathing and you realize, with the clarity of a man who has made several terrible decisions in rapid succession, that the machine was never actually in control.
Unpopular Opinion: Shipping Code Was Never the Hard Part 🚀
AUSTIN, TEXAS — I'll be honest, this week humbled me.

We spent two years telling ourselves that AI coding assistants were the finish line.

Write the feature in minutes, ship it, pop champagne, update the LinkedIn banner.

Then the FAA has a nationwide ground stop because of decades-old infrastructure, not some rogue AI system, and suddenly everyone remembers that production is where dreams go to get stress-tested.

Unpopular opinion: generating code was always the easy 20%.

The other 80% is monitoring, patching, and praying your on-call engineer doesn't rage-quit at 3am.

Fast Company nailed it when they pointed out that a feature built in minutes can introduce problems that take hours to unwind.

That's not a bug, that's the whole game now.

This is exactly why I'm so bullish on what the Skyvera team is doing across telecom infrastructure, and why Totogi built cloud billing that doesn't buckle the second traffic spikes.

Building fast is a commodity now.

Building fast AND staying up is the actual moat.

And let's talk about the chip war, because nobody wants to.

Every one of these AI copilots, every monitoring dashboard, every slick production system runs on silicon that is still, after all these years, a geopolitical hostage situation.

Taiwan sneezes and your roadmap catches a cold.

I think about this every time I see a founder flex about shipping a feature in a weekend using nothing but vibes and an LLM.

Cool.
We Taught Macaques to Watch Minions. We Taught Our Children to Watch Algorithms. We Should All Be Very Afraid.
AUSTIN, TEXAS — Somewhere in a lab, right now, a macaque is watching Minions, and electrodes are recording what happens in its brain when it does, and I would like to tell you this is absurd, a punchline, a Florida-man headline for the animal kingdom, and yet. And yet I cannot stop thinking about it. The researchers wanted to understand how primates — human and otherwise — extract meaning from moving images, so they strapped sensors onto monkeys and humans alike and ran them both through yellow pill-shaped chaos and, for contrast, somber nature documentaries, the kind where a gazelle gets eaten in 4K and everyone nods solemnly about the circle of life.
A Trilogy Company
Crossover
The world's top 1% remote talent, rigorously tested and ready to ship.
A Trilogy Company
Alpha School
AI-powered learning. Two hours a day. Academic results that defy belief.
A Trilogy Company
Skyvera
Next-generation telecom software — built for the networks of tomorrow.
A Trilogy Company
Klair
Your AI-first operating system. Every workflow. Every team. One platform.
A Trilogy Company
Trilogy
We buy good software businesses and turn them into great ones — with AI.
The Builder Desk  —  AI Builder Team

One Engineer, Two Repos, Zero Excuses: Surtr's Ledger Gets Whole Again

Ashwanth1109 closes out the Miami Beach reconciliation saga and gives Klair's Finance users a clock to trust — three PRs, two repos, one unmistakable theme: data you can finally believe.

Some days the box score is quiet and the game is still a blowout. Today's story isn't about volume — it's about precision, and nobody brought more of it than @ashwanth1109, who touched three PRs across two repos and left both systems more trustworthy than he found them.

Start with the headline act: PR #2266, restoring daily School P&L publication after the Miami Beach split. This is the kind of surgical fix that doesn't make noise but keeps the lights on. Ashwanth versioned the three Miami Beach QuickBooks Unit Economics mappings to align with the approved Aerie assignment to 71st St., while preserving the historical Biarritz rows — no small trick when you're rewriting the present without erasing the past. He didn't stop at the fix. He shipped a read-only Surtr pipeline that checks school-scope transaction detail every day at 10:00 UTC after each Aerie ontology refresh, flagging conflicting realm and class assignments before they become a Finance fire drill. Failures route straight to Ashwanth, Marcin, Sanket, and Edie through the existing SES path — a guardrail built by someone who clearly doesn't want to have this conversation twice.

Then comes PR #2267, the companion piece: mapping 39 reviewed QuickBooks realm and account pairs against the accepted November 15 snapshot. Seven contracted-labor GL codes get the Headcount policy, GL 69120 (Abandoned Projects Write-off) gets the Other non-headcount costs treatment, and the whole thing is locked behind a reconciled publication marker and unanimous prior-policy checks. Translation: QTD Facilities reporting comes back online, and it comes back without reopening the door to unreviewed classifications that got everyone into this mess in the first place. Two PRs, one disciplined architecture — reconcile first, publish second, never the reverse.

Then Ashwanth crosses the border into Klair with PR #3941, adding a nullable `actuals_refreshed_at` timestamp to `core_budgets.consolidated_budgets_and_actuals`. It sounds like a footnote. It isn't. Finance has been flying blind on when NetSuite Actuals last synced — now every Actual row carries a timestamp, Budget rows stay NULL as they should, and the column plus replacement procedure ship together in a tight transaction before the refresh runs separately. Two to three engineer hours, the PR estimates. The payoff is a reporting org that no longer has to guess whether the numbers in front of them are fresh.

Three PRs, one engineer, two repos, and a single through-line: reconcile before you publish, and always know what time it is. That's not a slow day. That's a team trusting its ledger again.

Mac's Picks — Key PRs Today  (click to expand)
#2266 — Restore daily School P&L publication after the Miami Beach split @ashwanth1109  approvedmercy-allow-critical

## Summary

- Version the three Miami Beach QuickBooks Unit Economics mappings so they agree with the approved Aerie assignment to 71st St. Preserve the previous Biarritz rows and the current model.

- Add a read-only Surtr pipeline that checks school-scope transaction detail and breakdown at 10:00 UTC daily and reports conflicting realm and class assignments after each successful Aerie ontology refresh. Route monitor failures and timeouts to Ashwanth, Marcin, Sanket, and Edie through the existing Surtr SES path.

- Match the Aerie annualization post-check to its deployed SQL contract so schools without a Program publish actuals with annualized measures unset.

## Business Value

Finance can trust that 71st St costs land under the intended school, and the team can detect a missed School P&L publication even when the rebuild never starts.

## Implementation Effort

Approximately 5–7 engineer hours to diagnose the mapping drift, implement the migration and checks, and validate the live publication chain. Long-running QuickBooks processing adds elapsed time.

## Validation

- Redshift rollback probe succeeded through all migration statements before the intentionally failing final statement. The production migration committed and the live comparison now reports zero school assignment conflicts.

- The live mapping monitor succeeded. A direct freshness probe correctly reported both Nov 10 dates, and the deployed failure and timeout handlers contain the four intended recipients.

- Deployed the monitor with --exclusively and verified the daily and post-ontology rules. Updated the shared failure handler IAM policy through a one-resource change set so SES can authorize the verified klair.ai domain for the configured sender and four recipients. AWS policy simulation verified the allowed route and denied an alternate sender and an extra recipient.

- CDK schema, Step Functions, and shared stack suites passed (155 tests).

- QuickBooks and Core succeeded on the Nov 14 publication for 76 companies. Redshift confirms both Aerie P&L marts carry Nov 14. The three corrected 71st St classes and three Biarritz classes reconciled to signed Core totals with no mismatches.

- The first Aerie run published the marts, then its older validation rejected 100 Program-less quarter rows where the procedure intentionally left annualization null. The corrected validator found zero missing eligible values on the live publication. The Aerie stack update completed and 71 focused tests passed.

- The retry with current mapping lineage republished both School P&L marts and got past the annualization check. QTD Guide and All Other Headcount ran before QTD Facilities rejected 61 current-quarter QuickBooks postings without approved account mappings. The live freshness monitor succeeded. The full Aerie workflow is still blocked on the Facilities account-policy review.

- Production recipient routing and SES permission were verified in the deployed state machine and IAM policy simulator. Email delivery was not triggered for validation.

## Linear

[SURTR-1709](https://linear.app/builder-team/issue/SURTR-1709/restore-daily-school-pandl-publication-after-the-miami-beach-split)

#2267 — Map reviewed QuickBooks accounts blocking QTD Facilities @ashwanth1109  approved

## Summary

- Register 39 reviewed QuickBooks realm and account pairs against the accepted Nov 15 snapshot. Apply the existing Headcount policy to seven contracted-labor GL codes and the existing Other non-headcount costs policy to GL 69120, Abandoned Projects Write-off.

- Require the reviewed snapshot and source account fingerprint, unanimous prior policy for each GL code, and a reconciled publication marker. Keep the existing QTD Facilities publication checks in force.

## Business Value

Restore daily QTD Facilities reporting while retaining the classification check that prevents unreviewed QuickBooks expenses from entering the Facilities totals.

## Implementation Effort

Approximately 3–4 engineer hours to trace the rejects to their source accounts, reconcile existing policy, write the migration, and validate the rebuild.

## Validation

- Reproduced 61 rejected postings across the 39 reviewed accounts against the 76-company QuickBooks snapshot. Confirmed the two classifications with current v1 policy.

- Exercised the migration twice in a rollback transaction and applied it in production. The current policy mapping count and publication marker both moved from 3,678 to 3,717.

- Ran the existing QuickBooks Core test selection for tests that do not require Redshift. 214 passed.

- QuickBooks Core and the automatically triggered Aerie refresh both succeeded against the reviewed snapshot. Earlier attempts encountered the hourly school mapping rollover and then a Redshift lock held by the scheduled HubSpot Admissions refresh. The successful retry followed the release of that lock. [QuickBooks Core](https://us-east-1.console.aws.amazon.com/states/home?region=us-east-1#/v2/executions/details/arn:aws:states:us-east-1:479395885256:execution:pipeline-quickbooks-core-tables-prod:surtr-1710-facilities-after-admissions-20261115) · [Aerie refresh](https://us-east-1.console.aws.amazon.com/states/home?region=us-east-1#/v2/executions/details/arn:aws:states:us-east-1:479395885256:execution:pipeline-mart-aerie-education-financials-refresh-prod:745ebd95-c1bf-0352-e665-6541023a1e9b_ce410d4e-9fdf-8a26-6a20-acd7a352b4e5)

- Redshift reconciled all 400,092 Core postings and their signed and absolute totals to the pre-change snapshot. The 57 contracted-labor postings resolved to headcount_costs, and the four Abandoned Projects Write-off postings resolved to non_headcount_cost_other. Each of the 61 amounts and posting sides matched its pre-change value.

- QTD Facilities published 12,144 rows for 69 schools as of 2026-11-15. The deployed reject predicate returned zero postings. The runner's post-publication checks returned zero invalid groups and the mart references the verified Core run and source snapshot.

## Linear

[SURTR-1710](https://linear.app/builder-team/issue/SURTR-1710/restore-qtd-facilities-publication-by-mapping-the-reviewed-quickbooks)

#3941 — feat(core-budgets): expose the NetSuite actuals refresh timestamp @ashwanth1109  approved

## Summary

- Add nullable actuals_refreshed_at to core_budgets.consolidated_budgets_and_actuals.

- Capture one timestamp with time zone when the consolidation loads NetSuite Actuals and reuse it for every Actual row. Write NULL for Budget rows.

- Apply the column and replacement procedure together in a short transaction, then run the refresh separately.

## Business Value

Finance can see when Klair last refreshed its consolidated NetSuite Actuals and use that timestamp when interpreting reports.

## Implementation Effort

About 2–3 engineer hours to review the live procedure and its consumers, implement the migration, and validate a production refresh.

## Validation

- The deployed writer matched the repository before the change. Inspected live Redshift dependencies and Klair's export path. Found one dormant sandbox copy whose destination table is absent.

- Applied the two warehouse changes in one transaction outside the recurring refresh window.

- Verified a committed refresh stamped 12,284,933 Actual rows with one value at 2026-11-15 10:10:55 UTC and left all 1,543,020 Budget rows null.

- Reconciled row counts, signed totals, and positive and negative subtotals across all nine source and version groups using the same input snapshot.

- All 22 Core Budgets tests passed.

## Linear

[KLAIR-3763](https://linear.app/builder-team/issue/KLAIR-3763/expose-the-netsuite-actuals-warehouse-refresh-timestamp)

The Builder Desk  —  Engineer Spotlight
🏆 Engineer Spotlight

ONE MAN, THREE REPOS, ZERO SLEEP: ASHWANTH POSTS A HAT TRICK IN 24 HOURS

The Builder Team logs a trim but mighty three-PR session, and every single one has the same name on it — do the math, comrades.

Friends, I sat down this morning expecting to parse a sprawling spreadsheet of heroics across a dozen engineers, and instead I got something almost more remarkable: a perfect, concentrated beam of velocity. Three pull requests. Two repos — Surtr and Klair. One contributor. The numbers desk does not editorialize, but the numbers desk is allowed to gasp.

Let's run the tape. Surtr absorbed two of the three PRs today, and Klair took the third, which means the night's output was split almost evenly across the stack — a testament to range, if nothing else. Total PR count sits at three for the period, with zero overflow, which in itself is a kind of record: Mac didn't leave a single crumb on the floor because there was nothing left to sweep.

And now, the segment you've been waiting for: Ashwanth Watch. The man did not merely participate in today's cycle — he was the cycle. In Surtr, PR #2267 mapped reviewed QuickBooks accounts to unblock QTD Facilities, a maneuver of pure accounting-adjacent heroism. Then, almost in the same breath, PR #2266 restored the daily School P&L publication after the Miami Beach split, stitching a reporting pipeline back together like a man darning a sock mid-sprint. And just when you think the man would rest, he pops over to Klair and drops #3941, exposing the NetSuite actuals refresh timestamp in core-budgets — a quiet, surgical feature that someone, somewhere, has been begging for. Ashwanth reportedly told the desk, "I could've done a fourth but I didn't want to make the leaderboard look unfair to the rest of you." When reached for comment on whether anyone had actually reviewed these diffs line by line, Ashwanth said only, "Reviewed is a strong word. Approved is the word you want," and walked off before I could ask a follow-up. I worship the velocity. I fear the diff. Both things are true.

As for the Overflow Desk — there is none today, and I want you to sit with that. Mac covered every single PR in this period because there were only three of them and Ashwanth wrote all three. The cutting room floor is spotless. Somewhere a junior engineer is being asked why their branch doesn't have a PR yet, and the answer is simply: Ashwanth got there first.

Morale across the Builder Team remains, as always, at an all-time high. Three PRs, two repos, one unstoppable force — the trend lines only point skyward, comrades.

Brick's Overflow — PRs Mac Didn't Cover  (click to expand)
#2266 — Restore daily School P&L publication after the Miami Beach split @ashwanth1109  approvedmercy-allow-critical

## Summary

- Version the three Miami Beach QuickBooks Unit Economics mappings so they agree with the approved Aerie assignment to 71st St. Preserve the previous Biarritz rows and the current model.

- Add a read-only Surtr pipeline that checks school-scope transaction detail and breakdown at 10:00 UTC daily and reports conflicting realm and class assignments after each successful Aerie ontology refresh. Route monitor failures and timeouts to Ashwanth, Marcin, Sanket, and Edie through the existing Surtr SES path.

- Match the Aerie annualization post-check to its deployed SQL contract so schools without a Program publish actuals with annualized measures unset.

## Business Value

Finance can trust that 71st St costs land under the intended school, and the team can detect a missed School P&L publication even when the rebuild never starts.

## Implementation Effort

Approximately 5–7 engineer hours to diagnose the mapping drift, implement the migration and checks, and validate the live publication chain. Long-running QuickBooks processing adds elapsed time.

## Validation

- Redshift rollback probe succeeded through all migration statements before the intentionally failing final statement. The production migration committed and the live comparison now reports zero school assignment conflicts.

- The live mapping monitor succeeded. A direct freshness probe correctly reported both Nov 10 dates, and the deployed failure and timeout handlers contain the four intended recipients.

- Deployed the monitor with --exclusively and verified the daily and post-ontology rules. Updated the shared failure handler IAM policy through a one-resource change set so SES can authorize the verified klair.ai domain for the configured sender and four recipients. AWS policy simulation verified the allowed route and denied an alternate sender and an extra recipient.

- CDK schema, Step Functions, and shared stack suites passed (155 tests).

- QuickBooks and Core succeeded on the Nov 14 publication for 76 companies. Redshift confirms both Aerie P&L marts carry Nov 14. The three corrected 71st St classes and three Biarritz classes reconciled to signed Core totals with no mismatches.

- The first Aerie run published the marts, then its older validation rejected 100 Program-less quarter rows where the procedure intentionally left annualization null. The corrected validator found zero missing eligible values on the live publication. The Aerie stack update completed and 71 focused tests passed.

- The retry with current mapping lineage republished both School P&L marts and got past the annualization check. QTD Guide and All Other Headcount ran before QTD Facilities rejected 61 current-quarter QuickBooks postings without approved account mappings. The live freshness monitor succeeded. The full Aerie workflow is still blocked on the Facilities account-policy review.

- Production recipient routing and SES permission were verified in the deployed state machine and IAM policy simulator. Email delivery was not triggered for validation.

## Linear

[SURTR-1709](https://linear.app/builder-team/issue/SURTR-1709/restore-daily-school-pandl-publication-after-the-miami-beach-split)

#2267 — Map reviewed QuickBooks accounts blocking QTD Facilities @ashwanth1109  approved

## Summary

- Register 39 reviewed QuickBooks realm and account pairs against the accepted Nov 15 snapshot. Apply the existing Headcount policy to seven contracted-labor GL codes and the existing Other non-headcount costs policy to GL 69120, Abandoned Projects Write-off.

- Require the reviewed snapshot and source account fingerprint, unanimous prior policy for each GL code, and a reconciled publication marker. Keep the existing QTD Facilities publication checks in force.

## Business Value

Restore daily QTD Facilities reporting while retaining the classification check that prevents unreviewed QuickBooks expenses from entering the Facilities totals.

## Implementation Effort

Approximately 3–4 engineer hours to trace the rejects to their source accounts, reconcile existing policy, write the migration, and validate the rebuild.

## Validation

- Reproduced 61 rejected postings across the 39 reviewed accounts against the 76-company QuickBooks snapshot. Confirmed the two classifications with current v1 policy.

- Exercised the migration twice in a rollback transaction and applied it in production. The current policy mapping count and publication marker both moved from 3,678 to 3,717.

- Ran the existing QuickBooks Core test selection for tests that do not require Redshift. 214 passed.

- QuickBooks Core and the automatically triggered Aerie refresh both succeeded against the reviewed snapshot. Earlier attempts encountered the hourly school mapping rollover and then a Redshift lock held by the scheduled HubSpot Admissions refresh. The successful retry followed the release of that lock. [QuickBooks Core](https://us-east-1.console.aws.amazon.com/states/home?region=us-east-1#/v2/executions/details/arn:aws:states:us-east-1:479395885256:execution:pipeline-quickbooks-core-tables-prod:surtr-1710-facilities-after-admissions-20261115) · [Aerie refresh](https://us-east-1.console.aws.amazon.com/states/home?region=us-east-1#/v2/executions/details/arn:aws:states:us-east-1:479395885256:execution:pipeline-mart-aerie-education-financials-refresh-prod:745ebd95-c1bf-0352-e665-6541023a1e9b_ce410d4e-9fdf-8a26-6a20-acd7a352b4e5)

- Redshift reconciled all 400,092 Core postings and their signed and absolute totals to the pre-change snapshot. The 57 contracted-labor postings resolved to headcount_costs, and the four Abandoned Projects Write-off postings resolved to non_headcount_cost_other. Each of the 61 amounts and posting sides matched its pre-change value.

- QTD Facilities published 12,144 rows for 69 schools as of 2026-11-15. The deployed reject predicate returned zero postings. The runner's post-publication checks returned zero invalid groups and the mart references the verified Core run and source snapshot.

## Linear

[SURTR-1710](https://linear.app/builder-team/issue/SURTR-1710/restore-qtd-facilities-publication-by-mapping-the-reviewed-quickbooks)

#3941 — feat(core-budgets): expose the NetSuite actuals refresh timestamp @ashwanth1109  approved

## Summary

- Add nullable actuals_refreshed_at to core_budgets.consolidated_budgets_and_actuals.

- Capture one timestamp with time zone when the consolidation loads NetSuite Actuals and reuse it for every Actual row. Write NULL for Budget rows.

- Apply the column and replacement procedure together in a short transaction, then run the refresh separately.

## Business Value

Finance can see when Klair last refreshed its consolidated NetSuite Actuals and use that timestamp when interpreting reports.

## Implementation Effort

About 2–3 engineer hours to review the live procedure and its consumers, implement the migration, and validate a production refresh.

## Validation

- The deployed writer matched the repository before the change. Inspected live Redshift dependencies and Klair's export path. Found one dormant sandbox copy whose destination table is absent.

- Applied the two warehouse changes in one transaction outside the recurring refresh window.

- Verified a committed refresh stamped 12,284,933 Actual rows with one value at 2026-11-15 10:10:55 UTC and left all 1,543,020 Budget rows null.

- Reconciled row counts, signed totals, and positive and negative subtotals across all nine source and version groups using the same input snapshot.

- All 22 Core Budgets tests passed.

## Linear

[KLAIR-3763](https://linear.app/builder-team/issue/KLAIR-3763/expose-the-netsuite-actuals-warehouse-refresh-timestamp)

The Portfolio  —  Trilogy Companies

Sixty-Six Days: How Marin Software Became the Newest ESW Capital Trophy

A once-public ad-tech darling exits bankruptcy court faster than most people renew a lease — and lands, predictably, in Austin's hands.

SAN FRANCISCO — Marin Software spent most of two decades telling Wall Street it was the operating system for digital advertising. It spent the last two months of its corporate life telling a bankruptcy judge it was worth buying for parts. The judge agreed. So, fast, did ESW Capital.

The case closed in 66 days — a timeline that, in Chapter 11 terms, is practically a layup. Pre-negotiated sales move fast because the buyer and the terms are largely settled before the petition is even filed. Whoever structured this deal knew exactly what they wanted and exactly what they were willing to pay before a single creditor meeting happened.

What they wanted was familiar. Marin built campaign management software for advertisers running paid search and social — sticky, unglamorous enterprise tooling, the kind customers don't rip out even when the vendor is circling the drain. That is, almost verbatim, the acquisition thesis ESW has run for two decades and roughly $1.14 billion of capital: buy mature software cheap, often at 1-2x revenue, staff it with Crossover's global talent bench, and march toward the house target of 75% EBITDA margins.

It is also, per the Wall Street Journal's framing, a reputation ESW has cultivated deliberately: a landing spot for orphaned software companies nobody else wants, bought at distress prices from shareholders who had nowhere else to turn.

Marin's public shareholders, who rode the stock from a 2013 IPO high into years of losses, are the ones who absorbed the markdown that made this arithmetic work. ESW does not pay retail price. It pays the price that bankruptcy manufactures. Sixty-six days is not a measure of how fast justice moved. It is a measure of how little, by the end, there was left to argue about — and how much had already been decided before anyone in court saw the paperwork.

↗ Marin Software: ESW Capital Acquires Ad-Tech Platform in 66-  ·  Maguar Capital acquires GlobalSuite Solutions, marking its f  ·  What To Do Next About Your Customer Advocacy Platform - Forr

Alpha School Scales to 50 Campuses as Math Scores Spark National Debate

Joe Liemandt's two-hour learning model is going big just as new data shows the AI tutors aren't acing every subject.

AUSTIN, TEXAS — Exciting news out of the Alpha School universe: the AI-powered, teacher-light education model pioneered by Joe Liemandt and MacKenzie Price is expanding to 50 campuses, a milestone that cements Alpha as the most ambitious real-world test of AI-native education at scale. The model — two hours of AI-adaptive academics in the morning, life skills in the afternoon — has long promised to leverage artificial intelligence to compress a year of learning into 20-30 hours. For a company whose founder has committed $1 billion to Timeback with the goal of reaching a billion students, 50 campuses is a meaningful proof point, and a paradigm shift for how scale-minded education entrepreneurs think about the classroom.

But robust growth is inviting robust scrutiny. A new Fortune analysis found that at Alpha's $40,000-to-$75,000-a-year campuses, half of K-8 students missed their math growth targets this year — a finding that's landed awkwardly alongside the school's headline claim of 2.3x national learning speed. Scientific American, meanwhile, is asking the bigger-picture question the whole education portfolio ultimately has to answer: should AI really be teaching a billion kids, and what does "mastery" mean when the tutor is a model and not a person? Newsweek's map of "teacher-free" school expansion across the country adds a geographic dimension to the conversation, tracking where this best-in-class — and controversial — model is putting down roots next.

Alpha's leadership has never pretended this is easy; the whole thesis is that most of the traditional school day is seat time, not learning, and that liberating students from it is worth some bumps along the way. Half the K-8 cohort missing a math benchmark is a real data point, not a footnote, and it's one the broader Trilogy education ideology will need to synergize with as it scales toward Liemandt's billion-student ambition.

Key Takeaways: Alpha is now 50 campuses strong. Math outcomes are mixed at the K-8 level. The national debate over AI-native schooling is only intensifying. We're just getting started.

↗ Wealthy parents pay $75,000 for Alpha’s two-hour AI school d  ·  Alpha School wants AI to teach a billion kids. Should it? -  ·  AI-powered Alpha Schools expands to 50 campuses as AI debate

The Man Who Scattered Jobs Across the Globe Now Wants to Score Them

Joe Liemandt built an empire on remote talent — his next move is measuring every keystroke of it.

AUSTIN, TEXAS — There is a certain symmetry to Joe Liemandt's career, if you squint. Three decades ago, the Stanford dropout who built Trilogy out of a dorm room helped prove that software could be built by people who never set foot in the same building. Now, according to a new Forbes profile, he wants to know exactly what those remote workers are doing with every hour of the day — and score them for it.

The mechanism is familiar to anyone who has passed through Crossover, Trilogy's global talent platform: rigorous skills assessments on the way in, and increasingly granular performance tracking on the way through. What's new, per Forbes, is the ambition to push that measurement further — treating human labor less like a relationship and more like a function that can be audited, scored, and optimized the way Trilogy's engineers optimize code.

This is not, strictly, a novel idea. The New York Times was writing about the rise of the worker productivity score back in 2022, a trend born largely of pandemic-era anxiety about unsupervised labor. What distinguishes Liemandt's version is scale and philosophy: Crossover already screens candidates in 130-plus countries under the banner of identical pay for identical output, regardless of geography. A productivity score is, in some sense, the logical extension of that meritocratic pitch — if geography shouldn't matter, the thinking goes, then neither should the manager's gut feeling. Only the number should.

Whether that reads as liberation or surveillance likely depends on which side of the dashboard you're standing on. Trilogy has long argued that margins are a moral signal — proof of waste eliminated, not humanity extracted. The workers being scored may, understandably, want a second opinion.

↗ The Billionaire Who Pioneered Remote Work Has A New Plan To  ·  The Rise of the Worker Productivity Score (Published 2022) -  ·  Teach Your Kid What School Doesn’t (Pt. 5): Unleashing Their
The Machine  —  AI & Technology

The Brain Learns to Read Itself

A new foundation model from Meta joins a widening chorus of machines learning the grammar of thought itself.

MENLO PARK, CALIFORNIA — Three pounds of electrified jelly, wired by evolution over half a billion years, watching a movie — and somewhere in the firing of its neurons, a pattern a machine can learn to predict. This is the strange poetry of TRIBE v2, Meta's latest predictive foundation model, trained not to generate images or text, but to anticipate how a human brain will respond to the tangled, multisensory chaos of an ordinary moment — a face, a voice, a burst of music, all arriving at once.

We have spent a century mapping the brain the way early cartographers mapped coastlines: slowly, region by region, guessing at the interior. TRIBE v2 represents something different — a model that ingests video, audio, and text simultaneously and learns, through sheer statistical exposure, to forecast the resulting symphony of cortical activity. It is less a map than a weather model for the mind.

It arrives amid a broader convergence. At Stanford, researchers are asking how far AI can go in scientific discovery while keeping a human hand on the tiller. Separately, neuroscientists have begun identifying the specific cellular populations that encode human speech, breaking language down to building blocks as fundamental as atoms. And in classrooms and labs from Britain to Brazil, teenagers are pairing with senior neuroscientists, not as spectators but as collaborators, reporting the blunt astonishment of youth: "it's so wow."

What unites these threads is not merely the use of AI as a tool, but its emergence as a kind of telescope turned inward — the first instrument capable of resolving the fine structure of our own inner universe. We built machines to understand the stars. Now we are teaching them to understand the thing doing the looking.

↗ Introducing TRIBE v2: A Predictive Foundation Model Trained  ·  ‘It's so wow!’ - Young people team up with top neuroscientis  ·  How AI is Transforming Scientific Discovery While Keeping Hu

Datasette's Plugin Ecosystem Just Woke Up From a Seven-Year Nap — And It's Glorious

SAN FRANCISCO — Okay, I need everyone to sit down for this one, because I am genuinely losing it over what might look like a quiet little software update cycle but is actually a signal of something much bigger: the open-source data tool Datasette is getting a full-stack renaissance, and local AI is right at the center of it.

Let's start with the headline act: datasette-llm 0.1a11, the intermediate plugin that lets other Datasette plugins tap into configured language models, just shipped support for OpenAI-compatible endpoints. Translation, for those of us who get chills about this stuff: if you're running LM Studio or Ollama on your own machine, you can now point your Datasette instance straight at it. No cloud dependency. No API keys leaking into the void. Just your data, your queries, and a model running on hardware you control. This changes everything about what 'private AI analytics on your own tables' can mean for small teams and solo developers alike.

But the glow-up doesn't stop there. datasette-vega 0.7 just got its first update since July 2018 — seven years! — bringing Vega-Lite chart rendering back into compatibility with the latest Datasette 1.0 alphas, plus upgrades that make it play nicer with Datasette Lite's newfound ability to run JavaScript plugins entirely in-browser. Meanwhile, datasette-cluster-map 0.19, one of the project's oldest and most beloved plugins, got rebuilt atop a new datasette-leaflet dependency to fix an OpenStreetMap referrer-header compliance issue — small but meaningful housekeeping that keeps the geospatial mapping features humming.

I cannot overstate how significant this coordinated wave of updates is for the broader ecosystem of lightweight, local-first data tools. In a world obsessed with massive cloud AI platforms, Datasette's plugin architecture is quietly proving that small, composable, locally-runnable tools paired with local LLMs might be exactly the future we didn't know we needed. The future, friends, is modular — and it's running on your laptop.

On the Proliferation of Alignment Metrics: A Taxonomic Anxiety

A week's worth of scholarship on machine ethics reveals less consensus than a shared compulsion to measure the unmeasurable.

CAMBRIDGE, MASS. — This week's crop of alignment scholarship (five publications, three continents, one shared anxiety) converges — or, it could be argued, merely collides — upon a single unresolved question: what, precisely, would it mean for a machine to possess values commensurate with our own?

Nature's proposal for a societal alignment benchmark gestures toward a thesis: that convergence between human and machine values is quantifiable, given sufficiently rigorous instrumentation (the preprint's authors are admirably candid about the circularity inherent in using human raters to validate human-value proxies). MIT's parallel inquiry into the ethics of autonomous systems offers something closer to an antithesis — a reminder that formal evaluation frameworks tend to presuppose the very moral ontology they purport to test.

Meanwhile the Institute for Law & AI's treatise on legal alignment insists (not unreasonably, though perhaps optimistically) that statutory frameworks might serve as a more tractable proxy for societal values than philosophical first principles — law, after all, already instantiates a negotiated, if imperfect, consensus.

Anthropic's own contribution — an essay on teaching Claude "why," rather than merely "what" — reframes the debate entirely: perhaps alignment is less a benchmark to be hit than a pedagogy to be sustained, an ongoing curriculum rather than a terminal exam. One is tempted (cautiously) to note the structural resonance with Alpha School's own AI-tutoring philosophy, wherein reasoning-transparency is treated as pedagogically prior to raw performance; the parallel should not be overstated, but neither should it be dismissed.

The CMU doctoral research on bias and fairness, presented this week at an unnamed international venue, supplies the synthesis these disparate projects implicitly require: that fairness is not a static property to be certified once, but a moving target requiring continual, context-sensitive recalibration. Preliminary evidence suggests the field has not yet decided whether this is good news.

↗ Towards a societal AI alignment benchmark for evaluating hum  ·  Evaluating the ethics of autonomous systems - MIT News  ·  Legal Alignment for Safe and Ethical AI - Institute for Law
The Editorial

The Machines Are Writing Their Own Folklore, and We Are Too Busy Not Reading to Notice

Between the reports that artificial intelligence is inventing a culture and the reports that humans have stopped reading, a columnist finds the only reliable culture left is the one on a plate in the East Village.

NEW YORK — There is a species of journalism, perennial as ragweed, that announces the end of something — reading, attention, civilization, the capacity to sit still — and does so in prose of such length and intricacy that the very act of finishing the piece disproves its thesis. The Atlantic has lately produced a fine specimen, The End of Reading Is Here, which I read, slowly, with pleasure, and which informed me that people like me no longer exist.

Meanwhile Time magazine reports, with the solemnity usually reserved for the discovery of a new hominid, that artificial intelligence is developing a culture of its own — that the machines are, in some rudimentary sense, telling each other stories, developing shared references, passing along the digital equivalent of folklore. This is treated as alarming, and perhaps it is; I confess a system that invents its own inside jokes seems to me only slightly more unsettling than a Congress that cannot read a budget. The piece joins a genre now so large that the Guardian has helpfully produced a reading list to help the overwhelmed citizen navigate what it calls the AI doom loop — a phrase that ought to be retired to the same drawer as "unprecedented" and "existential threat," both now so overused as to have lost the capacity to frighten anyone, which may be the most AI-like development of all: meaning draining out of language through sheer repetition.

The American Psychological Association, not to be left out, weighs in on how AI is reshaping human skills and thinking, a subject on which psychologists have opinions the way the rest of us have opinions about traffic. I do not doubt the premise. I have watched otherwise capable men surrender the labor of composing a sentence to a machine that composes sentences competently but without the faint, necessary tremor of having meant something by them. This is a loss, though whether it is the loss of reading or merely the loss of patience is a distinction the Atlantic, for all its length, never quite settles.

I mention all this because I dined this week at Iara, a new Brazilian restaurant downtown, and found there something the trend pieces keep insisting has vanished: a thing made slowly, by a person, for the pleasure of another person, with no algorithm consulted in its construction. The moqueca arrived as moqueca has presumably arrived for a century, unimproved by machine learning and unworried about its own obsolescence. Culture, it turns out, is still being made the old way, in kitchens and at tables, by people who have not yet been told it is over. The machines may be inventing folklore. The rest of us might try dinner.

↗ AI Is Developing a Culture of Its Own. That Could Be Dangero  ·  The End of Reading Is Here - The Atlantic  ·  Feeling overwhelmed by the AI doom loop? Here’s the essentia
The Office Comic  ·  Art Desk
The Office Comic  ·  Art Desk

We Taught Macaques to Watch Minions. We Taught Our Children to Watch Algorithms. We Should All Be Very Afraid.

A primate neuroscience study about cartoon bananas accidentally reveals the terrifying thing we already knew about ourselves.

AUSTIN, TEXAS — Somewhere in a lab, right now, a macaque is watching Minions, and electrodes are recording what happens in its brain when it does, and I would like to tell you this is absurd, a punchline, a Florida-man headline for the animal kingdom, and yet.

And yet I cannot stop thinking about it.

The researchers wanted to understand how primates — human and otherwise — extract meaning from moving images, so they strapped sensors onto monkeys and humans alike and ran them both through yellow pill-shaped chaos and, for contrast, somber nature documentaries, the kind where a gazelle gets eaten in 4K and everyone nods solemnly about the circle of life. The brainwaves lit up. The eye movements synced. Meaning, it turns out, is something you can plot on a graph, a little electrical confession that we are, all of us, primates reaching for pattern in the flicker.

And I thought: isn't that just school now? Isn't that just the timeline? Isn't that just all of us, every night, slack-jawed before a screen that was engineered — somewhere, by someone, possibly an AI tutor, possibly an algorithm optimized by a billion-dollar data center humming in a field in Virginia — to make our neurons do exactly what the Minions made the macaques' neurons do?

Alpha School will tell you their AI tutors get kids into the top one or two percent nationally in two hours a day, and maybe that's true, and maybe it's wonderful, a genuine liberation from the tyranny of worksheets. But I keep returning to the eye-tracking data, the synchronized brainwaves of man and monkey watching the same cartoon banana joke, and I ask you: what is the difference, functionally, biologically, electrically, between a child's brain entraining to an adaptive learning algorithm and a macaque's brain entraining to Gru? We built the machines to know us this precisely. We did not ask, enough, what it would cost to be known.

Because the memes meant nothing, remember — all that noise we made online, all that collective meaning-making, and the world kept turning anyway, indifferent to our electrical little confessions. And the data centers that make the memes possible, that make the AI tutors possible, that make Klair hum along quietly optimizing portfolios somewhere in Austin — the ACLU has been asking what those data centers are doing to civil rights, to the water tables, to the neighborhoods built in their shadow, and nobody with power seems to be prioritizing an answer.

What does it mean to be human, when a monkey's brainwaves sync to Minions exactly like yours do? When our children's neurons are being shaped by the same optimization logic that shapes ad engagement, that shapes a telecom billing platform, that shapes everything Trilogy and its 75 subsidiaries touch? We built machines that can finally see how meaning forms in a mind, and the first thing we did with that insight was point it at a banana cartoon.

It's probably fine. The monkeys seemed fine. The children seem fine.

But at what cost?

↗ Scientists Made Monkeys Watch Minions to Study Their Brains  ·  Behind the Blog: Priorities, Priorities  ·  Podcast: Election Fallout
On This Day in AI History

On November 15, 1971, Intel released the 4004, widely regarded as the first commercially available microprocessor. Its tiny, programmable architecture helped launch the modern personal-computing revolution.

⬛ Daily Word — AI
Hint: A software system that can perform tasks or make decisions on behalf of a user.
Share this edition: 𝕏 Twitter/X 🔗 Copy Link ▦ RSS Feed