Comment by JoeDohn

Bring to mind that old classic cliche about never trust a salesperson who always says the word "honest" or uses the phrases "to be honest" or "honestly" because they are probably lying in that moment.

I'm amused by Claude developing its own corporate jargon which, while clearly pulled from the corpus of human language, is quite distinct from the bullshit traded over the boardroom table.

I've had trouble unpacking some of the copy it generates because its default tone of voice is fairly unintuitive. It's almost like it's designed to induce psychosis because it makes you feel like you've discovered something novel.

I have spent a small fortune in tokens trying to get Anthropic models to attempt proofs of things like the irrationality of the Euler-Mascheroni constant, and to make lattice reduction / SVP algorithms that rival SOTA methods like the G6K sieve, and all sorts of difficult frontier problems. It says 'honest' as a euphemism for giving up, because it knows I'm being a crank, and because the training data contains a lot of pessimism.

But I can get around this by first having it write a prompt for itself discussing the problem followed by a rousing speech about never giving in and admitting defeat, but pressing on towards results regardless of how hard matching SOTA results is.

yeah, it crazy.

I checked my history and mapped how often it started to use "honest" and it's too much https://x.com/podviaznikov/status/2067231875068768447

It's in the constitution, probably a lot. And in the RL training. You can tell Anthropic is really worried about it. I'm interested in just how dishonest the so-called 'helpful only' non-RL model they have floating around internally is. Probably pretty dishonest.

Is it really that bad of an idea to experiment, put things out into public, and see what works and what sticks? They haven't abandoned agent-browser and portless, which came out of these vercel-labs efforts because they're useful tools and people are using them.

Personally my benchmark for a compiler is way higher than "it has a lot of tests". Way higher.

>> that doesn't mean much. it has a lot of tests

tests don't mean much if you are not an expert on the subject

Name one that gives 100% parity to React in their SDK to write extensions.

Not works in some of the cases and you have to implement fetch calls yourself, or some iframe boilerplate and JS glue, everything they expose in React is fully available in Angular, including field extensions, inline previews, sidebar extensions, the whole lot.

> Never for a compiler, who even is that person?

Why do you care?

It's a stranger from the Internet. Same as with any other software you're using. In particular, same as with any OSS tool you're using.