
What happened
In a Latent Space episode, Anthropic Claude Code team engineer Thariq Shihipar answered a host's question by calling CLAUDE.md something he thinks disappears at the limit, saying it may be better to start new projects without it.
Why it matters
A builder of a file that Claude Code users touch daily is hedging against it, and his advice to start without it is something developers can act on immediately, not just a prediction.
What to watch
The claim is hedged as "not quite true yet" and "in the limit," so the test is whether Anthropic or Claude Code users begin dropping the file in practice; he also flagged AskUserQuestion and Auto Mode in the same conversation.
WHO IT HITSClaude Code users and developers who maintain CLAUDE.md files to steer the tool's behavior will want to weigh dropping it on new projects, though the source himself frames this as a future state rather than current advice.
Summaries like this, in your inbox every morning.
The episode sits at an unusual intersection: the person discussing CLAUDE.md works on the Claude Code team, so the remarks carry weight beyond commentary. The write-up points out that the host introduced him as the person who first added the AskUserQuestion tool, and that a previous episode's Matt Pocock had already named him as a source of Claude Code knowledge, giving this conversation a thread back to earlier material.
That closeness is also why the write-up is careful. It notes that "CLAUDE.md will disappear" and "smart models will get cheaper" both came with qualifiers like "in the limit" and "not quite true yet" in the original video, and it deliberately kept those qualifiers in titles, subtitles, and descriptions. The stated reason is to avoid inflating an employee's personal statement into an official company position.
How much weight to give the CLAUDE.md remark likely depends on whether it reflects an internal direction or a personal view about where coding agents are heading. The write-up itself frames the bet as an experiment, pre-registering which clip it expects to do well and what result would count as evidence against its assumptions.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
ElevenLabs completed a $300 million employee tender at a $22 billion valuation, and co-founder Mati Staniszews…

On October 15, OpenAI announced the Rosalind Academic Plan, giving approved US university labs free access to…

A Qiita article compared automated red-teaming methods for LLM apps — GCG, PAIR, and TAP — using mock implemen…

Liquid AI released d1-ultra on October 15, 2026, an upgraded open-weights decision model that averaged 76.7 ac…

Circlia started a proof-of-concept for TASTE SIGNAL: visitors to Holland Village Cafe FUKUOKA taste cookies an…

In a webinar hosted by French Mathematical Society president Isabelle Gallagher, Fields Medal winner Hugo Dumi…
