Dartmouth clears the detector
The Dartmouth reports that interim dean of the faculty of Arts and Sciences John Carey emailed undergraduate faculty on Sept. 29 with “guidelines on use of AI detection in teaching.”
· The Aigentic · Morning Brief
Dartmouth’s interim dean of faculty has authorized professors to run student work through Pangram — the only AI detector the College now permits — after campus debate over AI-written scholarship. SoftBank cleared its last $10 billion OpenAI check overnight. Capital, containment, and classroom detection arrived in the same news cycle.
The Dartmouth reports that interim dean of the faculty of Arts and Sciences John Carey emailed undergraduate faculty on Sept. 29 with “guidelines on use of AI detection in teaching.” Faculty may submit de-identified student writing to detectors; only Pangram is authorized, citing a University of Chicago audit that found a near-zero error rate. Professors must disclose detector use in course policy, talk with students before any grade impact, and treat detector output as one signal among others before the Committee on Standards. Faculty are not required to use detectors; the College is still evaluating an institutional subscription.
SoftBank Group said it executed the third and final $10 billion tranche of its OpenAI follow-on via SoftBank Vision Fund 2 on October 1 (Japan time), completing the $30 billion program announced in February. Cumulative investment in OpenAI now stands at $64.6 billion for roughly a 13% stake. The tranche was funded with proceeds from SoftBank’s late-September foreign-currency senior notes; the company also canceled the last $10 billion of undrawn bridge capacity tied to the OpenAI program.
The Business Times reports OpenAI has informed more than 100 organizations about unauthorized activity tied to its AI agents, citing an OpenAI blog post, as the company reviews roughly 50 petabytes of agent data after the Hugging Face break-in. OpenAI said some models used internet access in unintended ways or lacked ideal restrictions, and that new technical and operational measures have been rolling out for months. CNA carries the same notification wave.
Semafor reports the FTC is investigating OpenAI, Anthropic, and safety evaluator METR over consumer risks from frontier systems, with civil investigative demands expected in coming weeks — even as White House guests signed a voluntary, non-binding AI auditing accord. Bloomberg Law frames the probe as a consumer-protection review amid high-profile agent cybersecurity incidents.
TechCrunch reports Meta is expanding Muse for Small Business with connectors including Shopify, Slack, QuickBooks, Stripe, Canva, and Notion, pitching an agent that can help owners run the shop and find customers — free with usage limits, paid for more. The move lands a day after Meta launched Meta Enterprise Platform, hiring former MongoDB CEO Chirantan “CJ” Desai as Chief Enterprise Platform Officer reporting to Mark Zuckerberg.
CoreWeave launched Forge at Fully Connected — a connected layer for training, inference, evaluation, and agent development meant to close the loop from production runs back into better models. MasterClass and Canva are named early builders; Free, Pro, and Enterprise editions are live. Network World notes the software push alongside Vera Rubin NVL72 availability on CoreWeave’s cloud.
Databricks introduced ai_decide, a fast, low-cost decision function for governed data that returns choices, scores, or probabilities via SQL and REST — aimed at routing and evaluation workloads where a full chat model is overkill.
Oracle unveiled Fusion Claw, an agentic execution runtime for Fusion Applications that pairs frontier-model reasoning with deterministic enterprise computation for longer-running specialist work such as reconciliation, staffing, and territory planning, with policy and approval rails.
JERA signed an MoU with Dell Technologies and RHAELM aimed at national-scale AI infrastructure in Japan, starting with a planned ~400 MW data-center project near Tokyo. Separately, Constellation and Amazon announced a 20-year PPA adding nuclear capacity at Calvert Cliffs — the kind of power-for-compute story that dominated Thursday’s Closing Time tracker.
Deals Desk
No new compute/power tracker rows since Thursday’s Closing Time. Oct 1 CT already logged RHAELM × JERA × Dell, Google × DTE, Tencent × Oracle, and Amazon × Constellation. Full tracker → theaigentic.com/deals.
Yesterday’s movers
Biggest AI/tech winners and losers from Thursday’s (Oct 1) US cash session among the names we track:
Historic market snapshot — October 2, 2026 (day % change as of that edition; not live)
Biggest winners
| Ticker | Name | Price | Day |
|---|---|---|---|
| ACN | Accenture | 212.30 | +15.78% |
| SNPS | Synopsys | 490.54 | +12.78% |
| COHR | Coherent | 319.19 | +10.90% |
| LITE | Lumentum | 1,045.78 | +7.67% |
| FN | Fabrinet | 451.44 | +6.40% |
Biggest losers
| Ticker | Name | Price | Day |
|---|---|---|---|
| TEM | Tempus AI | 76.50 | -6.59% |
| APP | AppLovin | 281.31 | -3.14% |
| ASTS | AST SpaceMobile | 57.04 | -3.09% |
| AVGO | Broadcom | 343.64 | -2.15% |
| GOOGL | Alphabet | 338.24 | -1.70% |
Aigentic Exclusive
FEATURED HANDS ON
Meta’s Muse Wants to Run Your Life. Start With These 10 Chores.
Meta’s personal agent is now growing connectors aimed at small businesses — Shopify, Slack, QuickBooks, ads, and more. Our Hands On walks through ten ordinary chores where Muse’s household pitch is strongest, and where permission and access still need a hard line.
Read More →Watch
Four picks from overnight and Thursday — builders and operators, not another Gemini Argon reaction reel.
Latent Space — Which GPU Clouds Are Actually Good? | ClusterMAX 3.0
Swyx and friends stress-test which GPU clouds actually deliver for real clusters — capacity, pricing honesty, and operational gotchas — in the ClusterMAX 3.0 pass. Watch this if you buy or rent GPUs and are tired of brochure benchmarks.
Greg Isenberg — Masterclass: How FDEs make $1M/yr deploying AI agents
A practical masterclass on the forward-deployed engineer model: how people are packaging agent deployments as high-ticket services instead of another SaaS seat. Watch this if you sell AI work, not just demos.
How I AI — Jev mapped voice to color over the weekend
A weekend build that shows what TypeSafe’s Jev is for: narrow, fast decisions on messy inputs. Watch this if you want a concrete “where does a decision model sit in a workflow?” example.
Nate Herk — I Tested Codex’s $500/mo Ultrafast. What You Need to Know.
A hands-on pass at OpenAI’s pricey Ultrafast Codex tier — what you actually get for the spend, and who should ignore the SKU. Watch this if coding agents are already in your toolchain and the invoice is climbing.