Dean Hamstead

Dean coding in Perl. Probably won't follow you. This account won't tweet or retweet politics.

[Dean Hamstead @PerlDean]

@LundukeJournal Seems odd not to even allow bug finding?

[Dean Hamstead @PerlDean]

This may come as a shock to many business leaders, but you can run LLM's on your own GPU infra and not pay per token.

[Dean Hamstead @PerlDean]

@LoopandPixels Seems like he low-balled them.

[Dean Hamstead @PerlDean]

@tanujDE3180 Strange that a "staff engineer" was indulging so much toil

[Dean Hamstead @PerlDean]

@esrtweet @tekwendell There is nostalgic value to it. But nowadays you're mostly replacing onboard peripherals (sound cards, network cards) - and nvme means you don't even have to be concerned with cable tidiness.

[Dean Hamstead @PerlDean]

@dhh Github only wants one 9. But people seem okay with it.

[Dean Hamstead @PerlDean]

@AccidentalCISO That's legit. Curious what the coax-ethernet is.

[Dean Hamstead @PerlDean]

@ausstockchick Prices will go down, rents will go up, a new equilibrium will be achieved.

Renters will need to cover all the expenses rather than owners taking a loss - which is then covered in part by tax offsets and then fully by the next buyer.

[Dean Hamstead @PerlDean]

@ausstockchick Obviously the media will publish clickbait in either direction. Both are self fulfilling though.

[Dean Hamstead @PerlDean]

This is fun. One click and your company is gone. In this case a supposed "all-in-one" cloud provider that is apparently all-in-gcp not actually in them?

[Theo - t3.gg @theo]

Do you know what's crazy about Google deleting Railway's account accidentally?

It's not the first time they've done it.

[Dean Hamstead @PerlDean]

@neetcode1 Whats the all-in-one part? all of it doesn't seem to be in one?

[Dean Hamstead @PerlDean]

@ID_AA_Carmack It does seem like the sort of problem that should be completely solved.

[Dean Hamstead @PerlDean]

@esrtweet macports?

[Dean Hamstead @PerlDean]

@ausstockchick And above asking offer on your sale?

[Dean Hamstead @PerlDean]

True.

[Nandini @N_and_ni]

Normal people use AI for help.

Developers use AI at 3AM like unpaid farm workers.

[Dean Hamstead @PerlDean]

@ausstockchick Seems like the expensive alternative to a Unit trust.

[Dean Hamstead @PerlDean]

Indeed.

[Chinmay A. Singh @chinmay]

Layoff has zero correlation with AI. The reason usually is excessive hiring in the past. When Elon trimmed Twitter, that was the start. Companies waited to see the results and then floodgates opened.

What further gave boost to that decision was higher interest rates. Stockholders stopped being OK with share grants at their expense.

But many of these laid off employees have a big struggle ahead. This is because past employers did not make them do real work. So such employees have lost touch.

[Dean Hamstead @PerlDean]

@Amank1412 Uptime doesn't matter any more.

[Andrej Karpathy @karpathy]

Good post from @balajis on the "verification gap".

You could see it as there being two modes in creation. Borrowing GAN terminology:
1) generation and
2) discrimination.
e.g. painting - you make a brush stroke (1) and then you look for a while to see if you improved the painting (2). these two stages are interspersed in pretty much all creative work.

Second point. Discrimination can be computationally very hard.
- images are by far the easiest. e.g. image generator teams can create giant grids of results to decide if one image is better than the other. thank you to the giant GPU in your brain built for processing images very fast.
- text is much harder. it is skimmable, but you have to read, it is semantic, discrete and precise so you also have to reason (esp in e.g. code).
- audio is maybe even harder still imo, because it force a time axis so it's not even skimmable. you're forced to spend serial compute and can't parallelize it at all.

You could say that in coding LLMs have collapsed (1) to ~instant, but have done very little to address (2). A person still has to stare at the results and discriminate if they are good. This is my major criticism of LLM coding in that they casually spit out *way* too much code per query at arbitrary complexity, pretending there is no stage 2. Getting that much code is bad and scary. Instead, the LLM has to actively work with you to break down problems into little incremental steps, each more easily verifiable. It has to anticipate the computational work of (2) and reduce it as much as possible. It has to really care.

This leads me to probably the biggest misunderstanding non-coders have about coding. They think that coding is about writing the code (1). It's not. It's about staring at the code (2). Loading it all into your working memory. Pacing back and forth. Thinking through all the edge cases. If you catch me at a random point while I'm "programming", I'm probably just staring at the screen and, if interrupted, really mad because it is so computationally strenuous. If we only get much faster 1, but we don't also reduce 2 (which is most of the time!), then clearly the overall speed of coding won't improve (see Amdahl's law).

[Dean Hamstead @PerlDean]

@dhh SONIC?