Anthropic moves its main code repository off Git as daily merges top 30,000

Anthropic says AI-assisted coding overwhelmed its old source-control system. Scion handles merges and history searches on servers, while Anthropic runs more tests after changes merge.

Read the full story →Listen to the edition

Anthropic ↗

Listen to this edition

Prefer audio? The briefing has chapters and a full transcript.

Follow on Apple Podcasts Follow on Spotify Transcript All episodes →

The briefing

The rest of the morning

4 more stories

GovernanceHot

AI labs dispute who can control outside safety reports

Reuters reported on October 13, citing two people, that Anthropic and OpenAI sought control over what evaluators could publish during talks. Anthropic disputes that this describes all of its work. The company says established evaluators control their findings, while Anthropic keeps approval rights for new firms. OpenAI says it does not seek approval over full reports.

The terms can change during negotiations. Reuters says some of Anthropic's discussions had changed by October 8. Both companies distinguish confidential information from findings. Amodei's September proposal allows narrow redactions but lets reviewers disclose important omissions. Reuters did not report that labs control every outside review.

Reuters via WTVB ↗Dario Amodei ↗

Read article →

Politics and regulationWarm

British MPs ask who gets early model access and who can stop a release

The UK Parliament scheduled former researchers, industry policy leaders, and government officials to appear at an AI session on October 13. The agenda asks whether testers get access early enough and who can stop a model release. September correspondence explains the concern. MPs asked about access to Claude Mythos 5.1, and AISI said no organisation outside the US had access when the model launched.

AISI's September reply also says it tested GPT-6 Astra before release, so access depends on the model. The agenda shows what MPs planned to examine, not what witnesses said. It did not grant new powers, establish a safety finding, or enact a reporting rule.

UK Parliament ↗UK Parliament ↗AI Security Institute ↗

Read article →

Economics and marketsWarm

IMF says weak AI profits could spread losses through finance

The IMF's October 13 financial stability briefing says markets could fall if AI investment or profits disappoint. It also flags rising bond yields and cyber incidents. Its severe stress test combines shocks involving AI, bond markets, and energy. The IMF says banks remain resilient overall in that scenario, though weaker banks and non-bank firms remain vulnerable.

The IMF assessed a risk and ran a stress test. It did not predict a crash. Pension funds, private credit firms, and leveraged hedge funds can each spread stress through the financial system in different ways. The report's earlier analytical chapters add context, but they are not separate findings from October 13.

IMF ↗IMF ↗

Read article →

Infrastructure and securityWarm

Microsoft's AI network agent stops for approval before making changes

Microsoft engineer Arun Malik described an AI agent for production-network incidents on October 13. The agent collects device state and telemetry. Ordinary software validates the model's output before acting on it. The agent has only the permissions of the person who started it, and each tool has separate authorisation and logs. A human must approve production changes. Microsoft grants trust separately for each procedure and type of action.

This account comes from Microsoft, not an independent reliability study. Even a plausible diagnosis does not let the agent change customer traffic. Microsoft gives no fleet-wide failure rate in the article. Its safeguards do not prove that autonomous operation is safe.

Microsoft Command Line ↗

Read article →

The watchlist

Signals that could change the read

Release failures, recovery times, and total costs.

Signed reporting terms and redaction disputes.

Transcripts and powers attached to model access.

Earnings, borrowing costs, and liquidity pressure.

Across the desks Warm

Faster code production raises questions about testing, publication rights, and permission to act.

SoftwareCompany operating resultsA higher merge count does not mean more useful software reached customers.

OversightPublication terms disputedA lab can give evaluators access while still limiting what they publish.

FinanceStress scenariosThe IMF tested what could happen under stress. It did not predict a crash.

OperationsBounded writesA correct diagnosis does not grant permission to act.

Editorial direction, not a financial index. Each signal is tied to this edition’s reporting.

Edition validated · 5 stories · 9 unique sources

About HotTea & our sources →