Michael Orenstein

@thsottiaux Can you add the ability to have a different model + reasoning_effort for spawn_agents_on_csv too?

Yesterday we launched subagents in Codex. Today we released GPT-5.4-Mini, which is SoTA in its category. Coincidence or genius move?

@thsottiaux When the model deletes a file and rewrites it, the codex UI should not show a 'full red diff' then 'full green diff' - it should show the actual diff.

What are we consistently getting wrong with codex that you wish we would improve / fix?

@thsottiaux The model seems to have an obsession with 'minimum work, fastest time-to-patch, time pressure' and nothing I can say in AGENTS.md or otherwise can align it to 'optimal and correct implementation'.

@thsottiaux Generally the model is lazy, I struggle to believe these multi-hour-long task claims? I constantly find myself saying "yes continue" - even with ExecPlans, clear instructions "continuing until completion".

@thsottiaux While 'fast' mode is appreciated, for non-time-sensitive asynchronous tasks, a 'slow' mode would be well received. Half the token cost for 'flex' style processing.

@thsottiaux In medium-to-long sessions containing questions, when asked to do an action after a question has been answered, it will perform some file reads then answer the previously answered question instead of doing the action, often with identical responses.

I wrote a bit about constrained decoding optimisations for large language models. michaelorenstein.com/blog/zero-entr…

@ctoofeverything @jarredsumner `this.constructor.constructor('console.log("oh look RCE", process.env)')()`

Bun and Node.js docs have conflicting info about the node:vm module. Bun claims it to be suitable for untrusted code execution, while official Node.js docs explicitly say in red that it cannot be used for untrusted code execution. So what’s up with node:vm in Bun? @jarredsumner

@danielkelly_io I would say it's a "human first product" more than it's an AI product. The time travel debugging is excellent. The 'inline watch variables' are ergonomic and significantly faster than console.log debugging. Their support has been responsive and solved bugs I reported within hours

Anybody using this "AI-Ready Test Runner" wallabyjs.com Or how about it's sister produce Quokka? Is it worth the price?

@tmikov I can't believe 'tab' is the default for AI autocomplete, all that muscle memory of language-server suggestions gone, replaced with a 400ms+ wait to read the AI suggestion. I don't see how anyone can get into an 'output flow state', constantly having to read those suggestions.

@tmikov I'm convinced the only reason people think Claude Code is any good is because it _says_ its code is good (sorry, "production ready and literally perfect") and it calls _you_ a genius every third line. It's just weaponised sycophancy.

At this point there is no doubt in my mind that Claude Code and similar tools decrease my productivity in areas which I am familiar with. It takes longer to guide the LLM to do the right thing than to do it myself. I also lose more time to retain context and to find subtle bugs. Waiting for the LLM encourages task switching, which kills focus. It is clearly and measurably a net loss. Let me repeat this: I can produce working code with Claude Code, but it takes longer than writing it myself and even then the quality is inferior. The frightening realization is that I still find Claude Code useful in areas that I don't understand well or simply don't want to bother with. That is almost certainly a case of Gell-Mann amnesia effect - the tool is producing crap in those areas as well, but I am naively hoping that it isn't. An even more frightening realization is that some employers may not understand this and might require AI usage.

@AndaristRake Yeah I was working on a compile-time validated 'schema' library, branding number literals / enums with metadata. In 'accessor functions' it's reasonable to select them out with control flow when they're not the same primitive type, but I guess at runtime that's kind of nonsense.

@KillcoDer given the brands are somewhat hacky... I'm not sure what should happen here really but it feeeels the branded version should be kept and the non-branded version could be dropped? im not entirely sure though

@AndaristRake I have another one! I'm trying to narrow a union of a literal string and a branded widened string type with a strict equality operator. Is this behaviour expected? (Narrow3 in the playground link) #code/C4TwDgpgBAQgTgQwHYBMoF4oG8oCNGoBcUwcArtAL4BQ1A9HVACoAWAlgM5SckvQcIAttADubYCyhIyg3BDhdkaDqTZIA5lw4iEYSCgB0tBlADKqjVAA24+QitQIARzL3xIKLQDGAeyQqpBDg4HxEARgwoAAo1MDJgYgByG2A7K0SoAB8oFTg1dQBKDAA+biQ44Ax0TGTbRHSoAH4yiqhiaSsralBIKAA5IJDwyIAlCGAyOCQmcAgAHh6IHwAzQODQsNKTACIUtO2sqTJO+kYoKAA9RtOoBC4IAA9IL1SUbz8ApEHQgCZImPK8WIUVqqXqGQAZLACCgitlcvkiuhSrF4lUantwU0Wmj2scuot+t8RH9MGMJlMZpAFrMVmshj8toxdnV7Adsh0uiZzlcbgBJKAiHzHNAsBAAN2gj2er2srIcwB8UDFkqgAGsIGBKuIuPglO9-JUvusRABmf6ohJQUFpDLZKIIyxQ+BKApIlGAyrVDHyjLNS1tI4nQkDE3msnjSbTWY0yB041DU1MqAssFsw4OizqKDOmFwoMOG4864mAVCkXKiVSp4QF4QNASHiK+mhEhK1P7Q6cg2fYkAFgtnqSmPsdui0lk8hz0Nd7pxXuq1pHDX9nsD3ZD-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-0FYh0l7cSv2Haxp74VdDXqY9w3EmEAAME1GWB1V-pOgHzVxS28Rlq0I+tvSiRs5Oo+5Mkk6TGFYVYBzU5957XbcIWPUAA" target="_blank" rel="nofollow noopener">typescriptlang.org/play/?ts=5.9.2…

@AndaristRake My assumption (before actually spending the day with checker.ts and the debugger, a humbling experience) was any context sensitive types would be deferred, then inferred 'left to right', as deep / nested as required.

@AndaristRake Naively I would assume people avoid relying on this inference, given it produces `unknown`.

@AndaristRake I think I found another one. I'm confused as to why having that dummy 'context' parameter seems to fix the inference, or annotating the ctx in the function before when the inference already gets it right. Does this one seem familiar to you? #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-JflGTFSs+jAAVKn5LoqP5Kn+UaKppQEGkaYEQeuKJbnBFBKAA6hwnrtHw8ASlk5bOHi0nCSSSiiQQ4kqfmxBULErCCMxF6snJt6CPeJxKUuqnntJYqaR+1CyjpYR6XsBl7EZEqATAl5IDFXJxXioXhRQ-JoOZiJ+tBhLBjZIyISsWASahsb+ZceHYaMS5-ANBH0dmpHvCMzFUVGI00bO40kbm6SxAkIgCNWFDMaxaBLhxCI+pZ5Com1DmegACqoYCxAJQm8SSIx9YF9IJSlBKXte8mVQ+yl7KtCRqalL7pRg2nfrlP75eshVvk2MllTeFU8iFYWCHV5lQRup3oiM53hPAVBkFd3S3Z5e4PYIT1UrSElvVJH3RUjCm-UlkXAxpRVaZlf0wzlkO-uqhmalzJkgcaWimrgjVQQGMGUxQAAiHXIesyIxsRfXzQCY1ZsteZkQWXyzZhA2LXrjE7aoC57exUirkd3E4ySit3S1B6K31BvvKjNXs4zsnMz9CWPhcyUMyxINcxlVBZRDAvQyEItwzEpXlQpvvo5I9WS+BQA" target="_blank" rel="nofollow noopener">typescriptlang.org/play/?#code/MY…

@AndaristRake Here you go: github.com/microsoft/Type… Yeah I found the behaviour surprising. Thank you for all of your help!

@KillcoDer but I'm not inclined to scope creep this PR ;v you could open a new issue about ur thing and I could then push a dedicated PR to handle just what you are asking for (if only tests would come clean)

@AndaristRake Oh, yes you're right, I missed that. Thank you. I removed all the const types from the issue since that's unrelated to the constraint vs default bug.

@KillcoDer This isnt strange, if u want to get the default type param then it makes sense it wouldnt be a readonly prop - ur default is just a Record

@AndaristRake have I found a TypeScript bug or is this a known limitation? The default of my class generic isn't being used in some cases: #code/PTAECcFMHMEsHsB2oDKkAuBXADgWAFAEigBCAhgM6SgDkU2846k4NoAxgDaUWgDusdAAtQ6PvA5IK6UNEiIWsdhQJkARtPBl2Mrj1KVIAJUgMmLADzspMgLIAaSYmmoAfKADeBUKGyY1nEqgALYYZAAmZOhkAFygtqAAvJ4AvqCU8d6+-oHsIWGR0SjsQpDBsahJqem8KAQpBERgACKKAG6Q4RCmjMzg-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-UDNzg0bcBoCSdxM0S3con3D1wDpdTNPSdLYLU0DwNoNRHTYODs3gnwqOoFSyHwxAA0QINLWSABVRAtwneB2t7E9FiQNZlR8EwsHARAJx4ixKJ4t0Yri1cGLoA9PWrWk6041xsswNsWGzRjmI5VrOKyXTlWa0B+IioSTVEpgJKkmSAnk1r2s67rwAqqAIiQM9KriGgatYHS9LAHxDINIdTPMmwrPG5BPKEJz7Vc9yHOtTAZGfKV-K0IL+jGWBODtCQL2oa9xkmV0wTsvolQXMBLmkSAIlAQRQFpu0XJZlgaCjPZcV4ORgyCOk4QBYE3RRZMoReRr12isp1uYBLID-P6Uu1JtMtAbLdf-PLAO2wriqqo20mAAAqG3wbIOqHeAFUGph1Cfd9v3-YMrCENhgPQ7DpCADl4G17H7tKKBecF6gpcUPILK0S50EcR5opc+nb24vDUElcpfuS3G+oGpAhpGz0xq3SbkJmzA5oWvClua1bNbozbharGtGCBA6jpO8AzqYgpWKurjlVupq-MewSqgksT3uk2TvpLtqOvL4NAe58IQZdDSqvvR8UjnsyQ-hw0TO98jJOCSnHT5Ae6z5W5rPRwZ4AJ5zqH1rjWYvBwpL35AoZyv9oBCCiq-PaLpkhpWbKiY67ZwCdiyjlACQFrZg1Khle2Ts8E0FdrBd2ntvbhyoaheG7NqH0MjtHagsc-Lx2oEuI8UBTwumxvnSYhcWpbzLl1A2-RK6DWGl1OuX8LDN1botTuYIxjwMOq2NB48LpTy3gdARD0BLPRXm9OYH0N4WB+jvER7kKony0j5C+lDA5GSNKOLcoBeqLECIsacQ57r9QkbXSM9cJq9RhKAOkvVtjixjOkRALoMJhJ2HEXqGDQBtHgLALoLZIAdHQZE74vA6SJN5ogEYlp8IpLSRk0A8T8KgCyTkocxAAAK6I1y-xkKtFyAApac0U-LMBcJqQQkACD3UIhHbJlgJyhMzBObMeSJYmwUDk+Z8TLKYGoHEMYVMRn4HuoY8Sxj15fQsAAGUcEYGEWQ6QWGmTqY2E4FnRLpKc0AAAyScaoAA+k5qmiELJs1+Oy1TRh+Lc1w9yuy-NBQUow7zPmgB+Y8tZAK6lAqoCKGpFAJk5LOdiKJPxxmTJDJckUPgUUbLJWi7ZGKprUrIDsggQA" target="_blank" rel="nofollow noopener">typescriptlang.org/play/?ts=5.9.2…

@AndaristRake Reverse mapped types to the rescue! Thank you so much, that's much less hacky. I wasn't paying too much attention to the mutability, but now that you point that out, that's also strange.

@KillcoDer the difference between the single case and the array one is basically caused by the same codepath that affects this: github.com/microsoft/Type…

@AndaristRake ahah, my design document for this API is littered with "we could do it this way but Andarist has a bug report + open PR from 2023 here"

@KillcoDer it feels familiar to the point I've expected to have a PR fixing this but I can't find it 😅