Anthropic's Champion Kit for Claude Code surfaces among engineers

One of the issues of integrating AI in teaching is that there’s a steep price-usefulness curve to AI plans. A basic ChatGPT plan is useless (counterproductive?); Claude Code or Codex would add value but useful plans are expensive. Allowing unfettered AI use gives students with more means a big edge.

I have been noticing recently how much claude code has this antipattern

Got 3 Claude Code 1-week Guest Passes to give away #BuildInPublic, give me a shout if you want one and will DM you the link

watching @flo-bit.dev give claude code full permission to live-read @stream.place chat and build features in real time was one of the most deranged things i've ever seen , but it gave me an idea

One of the things I hate most about how Claude Code defaults to generating python unless you convince it not to is that it leaves copious docstrings on everything. These *look like* documentation but they're really just the LLM leaving notes for itself.

Finally got around to bundling the skills I'm using every day into a repo and Claude Code marketplace so anyone can install them. I'm still rebuilding from what I was using daily at work, but we'll get there.

Interesting new Claude Code feature: "Pasting a PR URL into the `/resume` search box now finds the session that created that PR (GitHub, GitHub Enterprise, GitLab, and Bitbucket)"

different situation because Sora was largely useless but this is a bet on Claude Code and token consumers multiplying token budgets

Which will be useful to them for about five minutes until Claude‘s “experiment“ with removing Claude Code from Pro packages has been deemed to have succeeded. The lack of educational pricing for leading AI coding models suggests they’re really on the edge of things provision-wise.

Last month, my friend Joey, a VP of Data Sci & all-around cool human, joined us on the #DSLab. He demonstrated the cool tools he uses, including how he basically doesn't type anything to #ClaudeCode anymore 🤷🏻‍♂️ which blew my mind. Between cmux, the 1Password cli, wisprflow.. I'm 🤯 youtu.be/t146uZt-dP4

I agree. Claude Code can do things I cannot even comprehend while still being dumb as a door knob. This is probably why discussing LLM tools in terms of intelligence is a fools errand.

Tip: If ur Claude Code token usage is burning up ur quota but you still want to use Opus, Opus 4.6 uses less tokens than 4.7 but is still pretty good Use this to start a new session: claude --model "claude-opus-4-6[1m]" Or this to switch in the middle of a session: /model claude-opus-4-6[1m]

Hit my Claude Max $100 plan 5hr session limit with Claude Code more than once today, but had a productive coding day on the Claude Code desktop app, pushing 158 million tokens in past 24hrs and max 2.63 million tokens/hour with 73% Opus 4.7 and 21% Sonnet 4.6 🤓

今日のAI関連記事 Claude Code の skill と agent を分けて RAG を実装した話 | Zennの「AI」のフィード Claude Codeを用いて複数ファイルにわたる実装を行う際、役割を全体の進行を管理するskill（オーケストレーター層）と、個別実装を担うagent（実装者層）の2層に分ける設計を推奨しています。 この分業により、RAGのような複雑なタスクも効率的に実装でき、再利用性も向上します。 このオーケストレーターと実行層を分離する構造は、API実装やデータ移行など、あらゆる複数ステップの実装に応用可能な汎用的なパターンです。

Who owns the code Claude Code wrote? https://legallayer.substack.com/p/who-owns-the-claude-code-wrote comments #legallayer.substack.com