Manuel Schipper

Staff AI Engineer @Snowflake. Opinions my own

[Manuel Schipper @SchipperAI]

A sad story on top of HN today - the surge of AI generated PRs means Ladybird will no longer accept PRs. AI is changing how OSS projects approach collaboration

https://t.co/GXrLJtuch6

"A substantial patch used to imply substantial effort, and that effort was a reasonable proxy for good faith. That assumption no longer holds."

[Manuel Schipper @SchipperAI]

Agent labs like @cursor_ai and @cognition are in a great position - they have proprietary data and are capable of post-training their own model on top of open models.

[Gergely Orosz @GergelyOrosz]

It's starting.

The company with the cheapest coding model at a good enough quality is set to win big, as devs + companies become more price sensitive.

Of all the major players, Cursor is in a v good position with their Composer model. Plus eg Factory with smart routing

[Manuel Schipper @SchipperAI]

@danshipper How do you all run goal 24x7 safely? What kind of sandbox or deterministic enforcement do you have in place?

[Manuel Schipper @SchipperAI]

@vasuman How are you all running goal unattended for hours safely? What kind of sandbox or enforcement have you set up?

[Manuel Schipper @SchipperAI]

the importance of understanding that being busy doesn’t mean being productive

“Sometimes the highest leverage move is to stop orchestrating entirely, close the laptop full of agents and just think hard about one single problem”

[Addy Osmani @addyosmani]

https://t.co/0DyFIXJueI

[dax @thdxr]

i have seen enough proof now that using a coding agent is a deep skill

it's confusing because the people you see heavily using them produce horrible results

but that's because it's a skill! you can get better and the ceiling seems pretty high - this is very exciting to me

[Manuel Schipper @SchipperAI]

don’t forget to touch grass

[Manuel Schipper @SchipperAI]

@plainionist I prefer to direct it to the parts of the codebase that matter to the task at hand selectively to preserve context.

[Manuel Schipper @SchipperAI]

@kobyjconrad Try running codex with https://t.co/hXzRq8xQja so you only get prompted for the non-safe stuff.

[Manuel Schipper @SchipperAI]

@adamshuaib Is this research public? Would love to read up on the details

[Manuel Schipper @SchipperAI]

great Sunday read

“Human communication is hard, and it is tempting to avoid it when you can sit alone with your clanker.”

[Armin Ronacher ⇌ @mitsuhiko]

Has been a while since I wrote about agentic engineering, so this time around some learnings of maintaining Pi as a junior maintainer to @badlogicgames :) https://t.co/TbD9Jvqk3t

[Manuel Schipper @SchipperAI]

@BHolmesDev @mattpocockuk this slaps hard 🤣

[Manuel Schipper @SchipperAI]

@PawelJLisowski @GergelyOrosz @badlogicgames @mitsuhiko You can prompt it to do so. Or you can set up autoresearch ala Kaparthy. For me personally, it means creating my own personal projects with plugin points and hooks so that others can build their own plugins without having to modify the core.

[Manuel Schipper @SchipperAI]

big takeaway from @badlogicgames and @mitsuhiko interview with @GergelyOrosz is that, just like you can ask Pi to extend itself, self-improving software is a paradigm worth following

thinking about how you can build primitives so agents can mutate projects without having to fork the repo

https://t.co/OlwnTGWiXE

[Manuel Schipper @SchipperAI]

@m_chirculescu btw, please do let me know if I understood your question. you may have been asking about Codex internals, in which case please DM me and I can answer there

[Manuel Schipper @SchipperAI]

Or try https://t.co/hXzRq8xQja for the smooth auto-mode feel of no permission prompts for safe stuff with zero extra tokens and latency.

[Boris Cherny @bcherny]

People often ask what my biggest tip is for getting the most out of Claude Code.

These days my #1 tip is: use auto mode

Auto mode means no more permission prompts. It is the key building block for multi-clauding: start a session, then while it runs, work on another session in parallel.

[Manuel Schipper @SchipperAI]

@bettercallsalva love the insight that fighting the mandate burns political capital. better response - work within the constraint

[Manuel Schipper @SchipperAI]

best thing I read this week: https://t.co/ZH392WRnKC

"If there’s an organizational mandate to leverage LLMs to minimize the time spent coding, then that’s a new constraint we can work with. We can figure out what good engineering looks like in that context. [...] It’s fundamental to understand, though, that this is not an individual’s or team’s call: it has to be an organizational decision"

[Manuel Schipper @SchipperAI]

to best answer this question, I built a short doc page: https://t.co/dA1UcyhDe6

tl;dr - you can build a nah preset and set policies to either allow or block (no ask). codex agents launched with this preset will run unsupervized and won't stop to ask, and dangerous actions are blocked

[Manuel Schipper @SchipperAI]

@emilheap @bcherny currently only the CLI. I'll add support for codex app to the roadmap