fhdsgbbcaA

That just means the bugs are so subtle you haven’t found them yet, they are there and unspooling the damage may be very painful.

Things I never want to hear about flight control systems before I board a plane: “if things compile there's a high chance it's correct”

.

If it pleases the stock market, it’s okay.

With 100% control of the stack and an insanely good emulator in Rosetta.

I think windows-on-arm is fairly instructive as to how likely RISC-V would go.

I think there is an LLM bubble for sure, but I’m very bullish on the ease with which one can generate new specialized models for various tasks that are not LLM.

For example, there’s a ton of room for developing all kinds of low latency, highly reliable, embedded classifiers in a number of domains.

It’s not as gee-whiz/sci-fi as an LLM demo, but I think potentially much bigger impact over time.

Must be my lying’ eyes, fooling me once again.

I’ll say the unholy combination of managing the python GIL, concurrency, and connection reuse is not my favorite topic.

Claim is LLM exhibit reasoning, particularly in coding and logic. Observation is mere parroting of training data. Observations trump claims.

If you don’t read or understand the code, then no, you aren’t reasoning.

The condition of “some people are bad at thing” does not equal “computer better at thing than people”, but I see this argument all the time in LLM/AI discourse.

Had to do with connection pooling.

I asked both OpenAI and Claude the same difficult programming question. Each gave a nearly identical response down to the variable names and example values.

I then looked it up and they had each copy/pasted the same Stack overflow answer.

Furthermore, the answer was extremely wrong, the language I used was superficially similar to the source material, but the programming concepts were entirely different.

What this tells me is there is clearly no “reasoning” happening whatsoever with either model, despite marketing claiming as such.

My first thought as well - “AGI via LLM” implies that our grey matter is merely a substrate for executing language tasks: just swap out bio-neurons for a few H100s and viola, super intelligence.

“Sixty percent of the time, it works every time”

Ok so if you are doing sentiment analysis of user product reviews you want to silently truncate emoji because you don’t like them? That’s a good idea how?

Good to hear they saw the light but after I switched to Postgres I never had a single regret.

In a competitive market where people make very long term engineering decisions based on stability and reliability you can’t fuck up this badly and survive.

That just means it was trained recently enough to have it in training data.

A database that doesn’t give you back what you put into it is never a perk. It literally can’t handle storing and retrieving the data.

Which you need to do exactly once.

It’s also faster to type.

UTF-8 is what made me switch. It’s insane MySQL has something called UTF-8 that isn't really UTF-8, but do have a type UTF8MB4 that actually is correct. This means if you use UFT-8 in MySQL, you can’t use emoji for example.

In my opinion we should focus on Xc.v4 running over Z1 hyper branches.

Except the “social contract” is manifested as laws and enforced by the state.

Currently at least 50% of online ads are outright illegal in most parts of the world.

Nobody is morally required to have their legal rights violated to get information. Period.

I’ll have you know IE6 is known to have caused numerous divorces.

Don’t give up, fight’s not nearly over.

There’s no argument to use chrome either. Hopefully they will lose users.

I think that’s a fine approach, but they are claiming Claude is a 100x coder which is patently absurd.

Data breaches, financial errors, and the like for an early stage company are a death sentence. That’s the type of error I’m suggesting, not some flakey CSS.

I mean systems critical infrastructure code where error tolerance is zero.

For one off scripts or sketching a concept quickly it’s good enough, and for language reference it’s generally useful.

However, one thing I’ve noticed with Claude in particular is it tends to overweight the top answers in stack overflow.

The problem there is top answers are rarely the best answer - rather they tend to be overly verbose and long, whereas the best answer is usually the second one that just tells you what function to call.

On multiple occasions I’ve had Claude answer a simple prompt with horribly verbose and complicated code.

Then I’ll say “what about this single call?” (Eg the type of SO answer than gets the second most votes), and it’s says “You’re right! That’s a much better answer”.

Likewise I suggest to anybody to take the domain you are most knowledgeable in and pepper your LLM of choice with lots of questions and see how much it knows.

You’ll get a feel for how much you can trust it in other domains - which is “not much”.