makeramen

Aren't Cursor Composer models like this too? At some point all the extra RL you do can be considered as proprietary information added.

Not suggesting this is right or wrong, but is sort of the nature of the technology.

I do however know a lot of successful engineers crediting programming on TI-8* graphing calculators to getting them into programming...

I wonder what the best modern equivalent would be? Perhaps web/JS is easy enough to get into these days?

Gemini does the same (though not with health data). The only way to opt out of training on your data is to disable all Gemini chat history.

Gemini is also the only major provider where you can't opt out of using your data for training without disabling chat history.

You don’t think openai, google and other companies’ employees don’t do the same thing?

Of all the companies where employees are ghost-contributing llm code, Anthropic would be pretty low on my list to ban.

Oh man, I ported this to the TI-89 back in 7th grade and made it slightly more school appropriate calling it “pop wars”, trading soda from different machines at different schools instead of drugs.

It will sometimes do this for gitignored files to avoid reading secret tokens in env files for example. But for certain languages that rely on code generation this can be a pain.

Ah I just started using ghostty maybe a week ago and I mostly use quick terminal so I might be missing the panes on the regular windows.

Zellij has been holding me over with tabs and panes (in quick terminal) while I wait for them to come to ghostty directly.

Tabs (and panes? I haven't tried yet) should work fine for regular terminal windows though.

The DoW wants to only be beholden to the laws, and not to Anthropics TOS.

So the question is: do you trust the government to effectively govern its own use of AI? or do you trust Anthropic's enforcement of its TOS?

I see it as trying to apply the bitter lesson to robotics. Specialized robots will always have their place, but humanoid ones can take advantage of all the design interfaces that already exist in the world for humans.

Similar to how claude code gained so much traction in terminal by just leveraging the command line interface that already exists for humans, no need to invent a domain specific MCP to just run shell commands.

I agree with you that it's far from the most efficient approach for specific tasks. But the analogy would be that you also generally don't want to use LLMs to do something you can "just" write a script for... that doesn't make LLMs useless though.

Both can be true. I have personally experienced both.

Some problems AI surprised me immensely with fast, elegant efficient solutions and problem solving. I've also experienced AI doing totally absurd things that ended up taking multiple times longer than if I did it manually. Sometimes in the same project.

> The fact that I can unlock and relock the bootloader is not a security issue or a risk. People who don't know what that means cannot possibly do it by mistake.

The second sentence is false. Lots of people blindly follow things and don't understand consequences until they brick their devices. Those who don’t break something won’t notice if they’ve silently backdoored themselves.

People asking for support after getting themselves into some weird hole they never should have been in because some friend or online article said so is super common.

Tailscale serve

That seems like a typo or incorrect info, the M5 MBP definitely can be configured up to 32 GB, and the Apple page mentions 32 GB explicitly as well.

You can probably infer some from their Ory case study: https://www.ory.sh/case-studies/openai

But given the option, do you choose bigger models or more reasoning? Or medium of both?

You don't do it manually. You have claude do it once you’ve guided it back on track to remind itself not to do it next time.

Seems like that was a preview model, unknown if this released version is different

Kotlin Native's choice to go with a GC over native memory management is my biggest issue with it and really limits its use for memory and performance sensitive use cases.

So basically [engineering] design is more important than implementation details.

I would say the "engineering" part of the design is also optional, as product design is also another lever of higher influence than code optimization.

That’s a recent development, used to be something else altogether.

I had a bunch of very tenured teammates that didn’t really know how to use git, so there were only a few of us comfortable enough integrating and interacting with an open source dependency repo.

I actually dislike that Ruby reads like natural language. The complex logic in large systems would take many paragraphs to describe in english but is often much clearer and more succinct in code.

I have yet to find a portable windows machine that doesn’t throttle performance on battery. Weird that no reviewers ever mention it but Macs have nearly no performance difference and they plugged in vs on battery and that’s a huge reason I don’t trust using any Windows laptops on the move.

You do keep your compensation if you don't move your address, otherwise you get paid a market rate adjustment for that area, which from what I've seen is quite fair.

> If you know HTML you JSX is very intuitive.

That's a poor assumption that appears to be based on your personal preferences and what you're comfortable with. I personally find both require some learning, but prefer the svelte version.

I don't use Signal to message strangers, only contacts who I have phone numbers of. Seems like that's the use case for most messaging apps like Whatsapp and iMessage that Signal was designed to "compete" with. Signal is a private messaging app, not an anonymous messaging app. A worthwhile distinction with very different (though overlapping) target audiences.

"you are not the target audience"

You can downshift most automatics manually or to a “low” gear for downhill. Not sure how many drivers are aware of it or are comfortable using it though.