anon

[anon @anonpat__]

@AmandaAskell It’s helpful to have historical scans available if you end up having symptoms. You can move to a clear diagnosis with lot more confidence

[anon @anonpat__]

@jordihays Congrats

[anon @anonpat__]

@CatNyanpital Developers. Vast majority just build crud apps. Rn pms write stories, we sprint, qa, ship. Tickets to code seems solved, devs just verifying and shipping. When even this is solved, I am puzzled what this looks like than

[anon @anonpat__]

@DavidSacks is this the anti-lawfare guy???

[anon @anonpat__]

@saltyAom This is because of “halting problem”

Erlang and Elixir probably have the best management of this

[anon @anonpat__]

@dwarkesh_sp @ilyasut King!

[anon @anonpat__]

@JeffDean This is an incredible release, Jeff. Kudos to entire team. One place I think it’s lacking is working through pure codebases compared to frontier models. It can’t code review like GPT-5. It’s great at HTML Canvas-style code, but can definitely do better on broader codebase work

[anon @anonpat__]

I think it’s annoying that I co-work with @AnthropicAI claude code and in so many instances the default commit behavior is it slaps its name on the commit. Something about this experience makes me feel really frustrated. Feels like LLM stealing from me

[anon @anonpat__]

@jaredpalmer @Microsoft @github @code @Azure Great win for Microsoft. Good hire

[anon @anonpat__]

@saltyAom Very cool, in the process of migrating my backend from Django to this, excited to see the velocity

[anon @anonpat__]

@emollick I think the products are shifting from gpt/gems to artifacts/canvas

People used to share prompts, now they will share custom built apps

[anon @anonpat__]

@steipete Agree, wouldn’t say it’s perfect. Claude is being dumbed, it still understands code navigation better I think. GPT-5 high is better at following the patterns in your code base, it doesn’t start its own random things, more stable. But might not understand your request as well

[anon @anonpat__]

@mitsuhiko I found the logging part the most useful. Haveing multiple processes and logging is a big challenge that cc always runs into.

[anon @anonpat__]

@scaling01 lol hard disagree 😂😂😂😂 what a delusional take

[anon @anonpat__]

@gdb I have tried CC, codex and Jules. CC is by far way more mature. Making pull requests for your KPI is a bad idea. I think the code I got from codex is far inferior to others

[anon @anonpat__]

@cognition_labs will you train on the data?

[anon @anonpat__]

@Galois_Capital I guess you know more than all the world's intel orgs. Dude got terra call right and now thinks he is GCR

[anon @anonpat__]

@simpsoka I tried it out and had it make a few commits.

It doesn't feel asynchronous when I have to go back every 5-10 min to say continue, please give me an autopilot mode where I can set the task and walk away. I don't want it to stop until it's done.

[anon @anonpat__]

@jarredsumner We need a vibe coding write up from you 👑, teach us sensei