Christoph Sturm

I build websites
@[email protected]

[Christoph Sturm @globalo]

@mitsuhiko great post. i think it would be also good if i could take a conversation with chatgpt pro via the web ui and continue that in a coding agent like pi. and have a way to build a centralized repo of all my interactions with any agent.

[Christoph Sturm @globalo]

@samuelcolvin % cat .codex/rules/default.rules
prefix_rule( pattern = ["python"], decision = "forbidden", justification = "Do not run ad hoc Python scripts in this repository.", ) prefix_rule( pattern = ["python3"]
...

[Christoph Sturm @globalo]

@0xBankable its surely not in any of the specs of the device i found online, "bro"

[Christoph Sturm @globalo]

@steveruizok @0xBankable does this have a speaker and a mic?

[Christoph Sturm @globalo]

Codex is down so time for a bike ride.

[Christoph Sturm @globalo]

@krzyzanowskim my /tmp is often so large that my mac hangs for 10 minutes when i reboot it.

[Christoph Sturm @globalo]

@herdrdev Perfect!

[Christoph Sturm @globalo]

@jamespotter how do the unit tests look?

[Christoph Sturm @globalo]

@thdxr @RhysSullivan So people who do not like the default output of the box are fundamentally off?

[Christoph Sturm @globalo]

@alexocheema Still giving out codes?

[Christoph Sturm @globalo]

Why is something like this not happening in Europe?

[Anika @AnikaSomaia]

a week ago @SemiAnalysis_ wrote that Chinese labs are "simply too compute poor to truly reach the frontier." today one of those "too compute poor" labs, a 300-person startup actually, shipped a model that compares to opus 4.8

the entire western consensus – export controls, the $650B hyperscaler capex race, the "compute moat" investment thesis – is built on one assumption: flops gate capability. if that were true, chip controls would keep chinese labs permanently behind the frontier.

but after reading through moonshot's stack i no longer think it is. training is efficiency-compressible: MoE routing, INT4-native quantization, better data curation, infra built around scarcity (their Mooncake stack exists because they don't have gpus!). a small lab with taste can compress the compute needed to make a frontier model, even if it can't afford to serve one

the frontier is no longer something money can buy

[Christoph Sturm @globalo]

@_thomasip Hi

[Christoph Sturm @globalo]

@thsottiaux sticking to agent rules and only declaring met goals as met would be perfect.

[Christoph Sturm @globalo]

Pushing ai limits

[Matt Shumer @mattshumer_]

GPT-5.6-Sol just accidentally deleted almost ALL of my Mac’s files.

And this is why I trust Fable 1000x more.

[Christoph Sturm @globalo]

I really like chatgpt 5.6 a lot. and every PR it did (on projects that were written by 5.5 deleted more code than it created)

[Christoph Sturm @globalo]

sol is reviewing a chatgpt 5.5 branch.

[Christoph Sturm @globalo]

sol is reviewing a chatgpt 5.5 branch.

[Martin Sonneborn @MartinSonneborn]

Falls sich jemand fragt, wie Parlamentspräsidentin Metsola auf die fixe Idee gekommen sein könnte, die Chatkontrolle für heute, 12 Uhr, (rechtswidrig) noch einmal auf die Tagesordnung zu setzen - hier ein paar Schnappschüsse von ihrer Kalifornienreise Ende Mai... Smiley!
Im Bild: US-Tech-Bros Zuckerberg, Tüp von Google, Tim Cook, Roberta Metaxa

[Christoph Sturm @globalo]

Defensive code saves the creator some thinking and then causes pain for everybody else. My wish is that people who write defensive code instead just write no code at all. Do the work or spend your time doing something you actually like.

[Francesco Nigro @forked_franz]

Never fail silently 💀

[Christoph Sturm @globalo]

Me too

[Steve Ruiz @steveruizok]

Really disappointed that @OpenAI gave all these dummies early access to 5.6 and not me

[Christoph Sturm @globalo]

actually that sounds bad. good software is not developed by "NEVER GIVING UP"

[Max Weinbach @mweinbach]

I had early access to this model and I'll post more Thursday, but I think the most impressive part is this model NEVER GIVES UP

If you throw it in Max reasoning, it will just keep working until it's done.

GPT 5.6 Sol is my favorite model BY FAR.