acoard

Hermes supports this too, just do `/handoff telegram` and your current session is moved to telegram, it pings you on your phone, etc. It's pretty nice, though I do still prefer the Codex dedicated app for remote control

Isn't that exactly what Plato's saying? The books cannot understand themselves, and we rely upon them, and in doing so that changes us.

Plato on how reading and writing make us more forgetful as we rely on this new technology:

> And so it is that you by reason of your tender regard for the writing that is your offspring have declared the very opposite of its true effect. If men learn this, it will implant forgetfulness in their souls. They will cease to exercise memory because they rely on that which is written, calling things to remembrance no longer from within themselves, but by means of external marks.

I think this works unironically. My mother is an avid gardener and can spend 8 hours a day gardening. When her life circumstances allowed for it, she hired a once a week gardener to do the tasks she didn't like (or had difficulties doing as a small woman), and still gardens the same amount. I've teased her for hiring a gardener, but she swears it's a huge help and boost to her gardening quality of life.

Executing people guilty of serious crimes is good and just. They should have a proper trial, and the crimes should be sufficiently serious, but execution is no more or less "barbaric" than the alternatives. As Adam Smith said, "mercy to the guilty is cruelty to the innocent."

What about longer rest periods? For example if I wait 1hr between sets I can do full weight again without dropping down weights with a 2-5min break. In fact I can get multiple more sets in and significantly increase my total volume if I spread a workout over a day (which is easier with WFH). Any thoughts on this? Is there not enough muscle fatigue with this approach?

> And LLMs slurped some of those together with the output of thousands of people who’d do the task worse, and you have no way of forcing it to be the good one every time.

That's solvable though, whether through changing training data or RL.

It's completely unreasonable. Calling someone a dickhead in any professional environment is unprofessional and should be grounds for dismissal.

Should bosses also get one free insult for their employees too? Obviously not.

Yes, the type system basically replaces a degree of unit tests. You no longer need unit tests that are basically just type checking. And it's more comprehensive too. It's much easier to have incomplete test coverage without knowing it than an invalid type system. With such a type system, the app will fail to compile. Of course, you still need unit tests for many other things, but the type system does remove a class of them.

Whether typechecking or unit tests is "better" is really a question of taste.

> We only have access to local models because they're subsidized too.

Yes, and the flow of future models may dry up, but the current local models we'll have forever.

I'm sympathetic to the issue of services getting worse, it sucks, but

> If an API delivers very solid results one day and crap the next and I spent a lot of money, how does that work? There are many people on reddit/youtube speculating why claude sometimes responds like a brilliant coder and sometimes as if it had a full frontal lobotomy. I see this in Cursor too.

This seems like an incredible over-reach. There's no predatory behaviour here. You're free to cancel at any time.

It's an incredibly fast moving field, a frontier field in software. To say that, in order to charge for something, you are legally bound to never make mistakes and have regressions, is an incredibly hostile environment to work in. You'll stifle growth if people think experiments might come with lawsuits unless they're positive it leads to improvement.

If they decided they were going to lock everything to gpt-2 and refuse to pay back any people who bought yearly subscriptions, sure I would be agreeable to considering this a bait-and-switch hoodwink. But that is clearly not happening here.

> Because memory bandwidth is the #1 bottleneck for inference, even more than capacity.

But there are a ton of models I can't run at all locally due to VRAM limitations. I'd take being able to run those models slower. I know there are some ways to get these running on CPU orders of magnitude slower, but ideally there's some sort of middle ground.

But it seems like Airbus is the exception to the rule, by and large. For example, the Typhoon Eurofighter has had disappointing exports.

> My next question is are there downsides to proof-of-stake, and why has bitcoin not moved towards it?

There are minor downsides in theory. Proof-of-stake is a bit less "democratic" and gives more voting share to those who have more money, vs. just are mining. In practice, proof-of-stake coins seem to be doing fine. The main one here is Ethereum.

The main reason Bitcoin hasn't is due to inertia and the interest for miners (who do proof-of-work) to use their democratic power to keep bitcoin invested in proof-of-work.