JMC

Indie Dev | Web & iOS

[JMC @jmc_dev]

I don't think most people using open weight models need them to beat opus, they just need them good enough for the small stuff at a fraction of the cost https://t.co/k1ZZMIIvni

[Theo - t3.gg @theo]

Terrible day to be an “open weight models are just as good as Opus 5.5” guy

[JMC @jmc_dev]

@expo Didn't expect expo to have duo support this quick, makes it way easier to just do it now instead of waiting until april

[JMC @jmc_dev]

@BrandonButch It's crazy that it starts at $2,899, most people would still be better off with an air and an ipad

[JMC @jmc_dev]

Seems like most people just go with whatever the agent suggests, I do it too when both options look fine https://t.co/Tlqgv1ySKz

[Alex Tamkin @AlexTamkin]

New Anthropic research: How AI agents shape experts' judgment calls

When experts delegate work to an agent, do they still make the key decisions? We tested this in a randomized trial with 50 professional data analysts.

[JMC @jmc_dev]

Even he has a human assistant because companies still want to deal with a person, I think that's what agents won't replace for a while https://t.co/tmILkJiTCK

[TechCrunch @TechCrunch]

Alexander Embiricos, the OpenAI executive responsible for dots, never wants to have a meeting with someone else's AI agent.
https://t.co/mX8JRflAn6

[JMC @jmc_dev]

Seems like nobody really knows what to do with these results yet, most people can't even check if they're right

[Theo - t3.gg @theo]

Never seen ChatGPT get so serious so fast before

[JMC @jmc_dev]

@MaxHirsch13 I feel like most onboarding gets confusing because it asks too many questions before you've even seen what the app does

[JMC @jmc_dev]

@levelsio I think apps stick around longer than people think, most people still don't trust an agent with anything important

[JMC @jmc_dev]

@kr0der It's just the codex computer use tools that are better, but you can ask claude to use codex computer use

[JMC @jmc_dev]

@petergyang Is terra actually better at coding or just faster? Seems like it was pitched more for computer use

[JMC @jmc_dev]

@levelsio I think the links are a big part of it, most people don't trust an answer unless they can see where it came from

[JMC @jmc_dev]

@cursor_ai I'd probably use this for small tasks and save opus 5.5 for the harder stuff

[JMC @jmc_dev]

@markgurman It's insane that the one iPhone with room for picture in picture doesn't have it

[JMC @jmc_dev]

@OpenAIDevs Hopefully it's fast enough to have codex test ios apps in the simulator, that's where most agents are too slow

[JMC @jmc_dev]

@ChadAppDev I thought they only remove ones that break the rules e.g. spam or reviews meant for a different app

[JMC @jmc_dev]

@theo Costs more than opus 5.5 on the lower settings and still scores worse, don't know why anyone would pick it at the moment

[JMC @jmc_dev]

Used Claude to create a product video for MomentClip

[JMC @jmc_dev]

@leerob To be able to have the chat for 2 bots up side by side

[JMC @jmc_dev]

@jacobrodri_ Apple must have Siri running app review

[JMC @jmc_dev]

@carlmonkft I think we underestimate how many people just don't care enough to cancel. Like I have talked to so many people that say 'i should cancel netflix I don't use it much' but they still keep it