Ci

Goldman Sachs is pulling AI coding tools including Claude Code and GitHub Copilot from traders and bankers, restricting access to engineers. The bank cited productivity and risk management, reversing its earlier public praise of 'vibe coding' by non-technical staff.

Enterprise AI • The Decoder

DeepSeek released a 'creator mode' letting DeepSeek Harness users build custom plugins, such as a 3D model renderer and a video editing workbench. The feature targets complex, task-specific workflows.

Product Launch • Techmeme

CCID reported that AI inference data in China has overtaken training data for the first time, with daily token calls above 175 trillion as of June—more than 1,000 times the early-2024 level, according to the firm.

Industry Data • Pandaily

GenRobot AI • Series A+

GenRobot AI, a Beijing-based company providing data infrastructure for robot development, has completed a $100 million Series A+ funding round. Participating investors include SAIC Motor, Malaysia's Eastworlds, Baidu, Shoucheng Holdings, and Galaxy General Robotics, alongside financial investors including GL Ventures, CITIC Goldstone.

ORCL

Oracle is selling $10B in common stock and $10B in bonds as it rushes to get financing for its data center buildout and chip purchases; ORCL falls 5%+.

OpenAI says its GPT-6 Astra model designed a hardware device for the Codex coding agent in under a week, handling enclosure, circuit board, and firmware. Several revision rounds were needed before production because the model didn't always account for manufacturing tolerances, and OpenAI says a newer internal model scores significantly higher on its engineering benchmark without sharing details.

AI Models • The Decoder

Revolut says it cut AI coding inference costs by roughly 90% in less than a year by making self-hosted GLM models its default over commercial tools like Claude Code. The move highlights a growing trend of enterprises replacing third-party coding assistants with in-house model deployments to control spend.

Enterprise AI • Techmeme

Based on a transcript excerpt: Fireship breaks down a series of major cybersecurity incidents, including a backdoor shipped in a Cursor extension, a hijacked script draining Polymarket funds, a reported $6.3B crypto theft from Binance, and an Anthropic demonstration where an unreleased Claude model reportedly gained root on unpatched Linux servers. The video explains the technical details and broader implications for developers and security professionals.

AI Security & News • Fireship

Stanford researchers tested 50 widely cited claims about LLMs against 20 current models and found 44% of the findings are already expired, with another 34% significantly weakened. The average lifespan of a claim is just 15.7 months, and claims about what models cannot do expire fastest.

AI Research • The Decoder

China plans to roll out broad AI safety standards covering development through deployment, targeting domestic AI model makers, according to Bloomberg. The standards come after global security incidents and signal Beijing's push for formalized AI governance.

AI Policy • Techmeme

SMCI

Super Micro reported Q1 revenue up 298% year-over-year to $19.95 billion, beating the $16.39 billion estimate, and forecast Q2 revenue above estimates. Shares jumped more than 7% after hours on the strong AI server demand.

PLTR

Palantir's quarterly results showed U.S. commercial revenue surpassing U.S. government revenue for the first time ever, a milestone that also contains a foreboding signal according to TechCrunch. The article does not specify exact revenue figures or stock movement.

Voters elected candidates favoring data center moratoriums and restrictive ballot measures in the 2026 midterms, dealing a blow to AI infrastructure expansion. NBC News polling found 76% of voters cited AI and data centers as an important factor, though exit polls ranked AI last among campaign issues behind corruption (90%) and cost of living (89%).

Policy • The Verge AI

DeepSeek reportedly plans to train its next-generation V5 model on Huawei Ascend chips, according to STAR Market Daily citing industry sources. The reported plan would extend Ascend's role from inference to the compute-intensive training of a new model.

Infrastructure • TechNode

Based on a transcript excerpt: The AI Daily Brief discusses a new MIT paper showing how a larger model can edit the weights of a smaller model, fueling talk of recursive self-improvement (RSI). It also covers Elon Musk's comments on the cost of GrokBot and Anthropic's updated language on RSI.

AI News • The AI Daily Brief: Artificial Intelligence News

Geoffrey Hinton told the Guardian it could turn out humans would have been better off never developing AI, warning the technology could trigger a large-scale incident this year or next such as a temporary blackout of financial or power systems. The Nobel laureate said he is not a 'doomer' about the technology.

Safety • Hacker News

AI industry PACs spent millions influencing the 2026 midterms with mixed results, and political groups are already looking toward the next election cycle. Separately, OpenAI's Chris Lehane is proposing a new social contract aimed at defining the 2028 election.

Policy • TechCrunch AI

fal • Series F

AI infrastructure company fal has raised $1.3 billion in a Series F funding round led by Bessemer Venture Partners, with participation from NVIDIA, Sound Ventures, HOF Capital and T. Rowe Price. The round doubles the company's valuation to $16 billion.

WULF

TeraWulf secured a credit backstop agreement with Google to support financing for its data center project with Anthropic in Kentucky, set to come online in H2 2027. The agreement provides financial backing for the project.

AMD → Taalas

AMD will pay approximately $2.9 billion in cash and stock for Toronto-based AI chip startup Taalas, with $1.4 billion dependent on unspecified conditions. The deal was announced in August.

Productivity

Datasette Apps 0.3a0 lets apps be tagged in Datasette Agent chat by typing #, which autocompletes from apps the user has permission to view; agent tools can add or remove stored queries an app may call, and granting a write query requires user approval first.

Productivity

datasette-llm 0.1a10 fixes a bug that affected the agent.datasette.io demo site, where the set of available LLM models was not correctly restricted to just those in the configuration.

Productivity

The newsletter frames AI agents as putting work within reach that once needed a team, and offers guidance on deciding what deserves doing, what your time is worth, and when to spend more; it cites Elon Musk saying his AI agent costs too much to serve for what it earns.