Hacker News: tjwebbnorfolk

New comment by tjwebbnorfolk in "One night in Uzbekistan: Why was this one data point so influential?"

tjwebbnorfolk — Sat, 22 Aug 2026 23:00:26 +0000

> had the axes cropped so you couldn’t see

Almost sounds intentional...

> This seemed indicative of a different issue, which is why we documented it in the comment.

Yea, that different issue is fraud.

New comment by tjwebbnorfolk in "A look under our trunk: what's in our compute"

tjwebbnorfolk — Fri, 21 Aug 2026 23:51:41 +0000

Maybe it's different where you live, but 95% of the buses I see driving around here are completely empty.

New comment by tjwebbnorfolk in "There's no such thing as a small software team anymore"

tjwebbnorfolk — Fri, 21 Aug 2026 03:12:04 +0000

There will, at some point, be a capex winter. Companies are splurging on a huge buildout and aren't going to buy GPUs at this rate forever.

That does not mean we're all going back to artisanally hand-crafting software. That's never going to happen.

New comment by tjwebbnorfolk in "Does anyone run Postgres without PgBouncer?"

tjwebbnorfolk — Sun, 16 Aug 2026 14:20:01 +0000

This is the first time I've heard pgbouncer mentioned in 15 years. I didn't know it was still around.

New comment by tjwebbnorfolk in "GLM-5.3: Frontier coding with emergent cyber capabilities"

tjwebbnorfolk — Fri, 14 Aug 2026 14:07:11 +0000

I was asking if this implies that 5.3 has the same number of parameters as 5.2. I can, in fact, read. What I didn't do is understand the implication of that statement. Thank you for your copy/paste service.

New comment by tjwebbnorfolk in "GLM-5.3: Frontier coding with emergent cyber capabilities"

tjwebbnorfolk — Fri, 14 Aug 2026 06:04:56 +0000

does this suggest 5.3 is the same # of parameters as 5.2?

New comment by tjwebbnorfolk in "Gemini 3.7 Flash"

tjwebbnorfolk — Thu, 13 Aug 2026 21:23:49 +0000

> especially given how hard it is to get an API key from them

What does this mean? Anybody can get an API key

New comment by tjwebbnorfolk in "Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models"

tjwebbnorfolk — Tue, 11 Aug 2026 14:34:31 +0000

They already announced it for release in April. Why am I being attacked here

New comment by tjwebbnorfolk in "Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models"

tjwebbnorfolk — Tue, 11 Aug 2026 06:40:59 +0000

It's already been trained. GDM said it was going to be released and it hasn't yet. Not sure why so much snark

New comment by tjwebbnorfolk in "Mark Zuckerberg attacks 'closed' AI rivals as Meta returns to open models"

tjwebbnorfolk — Tue, 11 Aug 2026 01:26:51 +0000

where's gemma 124b-a15?

New comment by tjwebbnorfolk in "Letter to Governor Abbott on responsible AI infrastructure in Texas"

tjwebbnorfolk — Mon, 10 Aug 2026 15:32:11 +0000

Nevertheless, experiments are necessary. Political entrepreneurialism is necessary if you want anything to improve.

You can experiment on everyone, or you can experiment at the local level. One is obviously preferable over the other.

Nobody has all the answers.

New comment by tjwebbnorfolk in "Why do we assume everyone should be working?"

tjwebbnorfolk — Mon, 10 Aug 2026 15:04:25 +0000

I mean, there's not like some "big silicon" conspiracy keeping people from homelabbing their own chips. I suppose we'll find out at some point, but it sure as hell isn't possible in 2026 AD.

New comment by tjwebbnorfolk in "Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows"

tjwebbnorfolk — Mon, 10 Aug 2026 14:40:48 +0000

More meta derangement syndrome on HN, what a surprise.

We all benefit when companies invest their resources in producing open models. No one thinks this absolves anyone of being terrible elsewhere. But we can still be happy about it.

New comment by tjwebbnorfolk in "Why do we assume everyone should be working?"

tjwebbnorfolk — Mon, 10 Aug 2026 06:54:20 +0000

Well, my ancestors would already say that our scarcity is effectively zero.

Yet we still have political arguments about healthcare and I get irritated when my AI service says I'm out of quota, two things that didn't exist 4 or 5 generations ago. Unless you consider whiskey and opium to be healthcare. It's amazing what we've adapted to.

New comment by tjwebbnorfolk in "2027 memory capacity is reportedly sold out"

tjwebbnorfolk — Mon, 10 Aug 2026 06:51:49 +0000

yea right now local models are like the swing producer. When claude throws 503s or 429s at me, or I run out of quota, I fire up a local model.

It's also resilience. I can still do AI stuff if the internet goes down.

New comment by tjwebbnorfolk in "Why do we assume everyone should be working?"

tjwebbnorfolk — Mon, 10 Aug 2026 06:45:16 +0000

> must attack things like the very concept of trillion dollar semiconductor fabs

Computers used to be the size of these fabs. Now we have them in our pockets. Not because we need to attack or rebel or something. It's because a lot of people worked very hard over many decades to make it possible. And it wasn't out of the goodness of their hearts -- it's because they had bills to pay and mouths to feed.

No matter how smart you are, it is not possible to create a GPU in one's basement. And no matter how holy you think you are, you need incentives to invest the time and effort to make this happen.

New comment by tjwebbnorfolk in "Why do we assume everyone should be working?"

tjwebbnorfolk — Mon, 10 Aug 2026 06:37:28 +0000

> Economics itself is the enemy. Economics arise out of scarcity.

Scarcity is a timeless fact of life in a finite world, it's not some abstract invention by "economics".

The problem isn't that more people want more things than they can have. The hardest possible life is that of a single human alone on a planet abundant in all natural resources. People and our inventions and our work and our ambitions to improve things for the next generation are the very things that reduce scarcity for other people we share this planet with.

New comment by tjwebbnorfolk in "Ask HN: What are you working on? (August 2026)"

tjwebbnorfolk — Mon, 10 Aug 2026 01:10:48 +0000

I put together a nationwide (US) land parcel dataset over the past year and a half, and have been building a bunch of apps on top of this data: https://landstats.com

New comment by tjwebbnorfolk in "2027 memory capacity is reportedly sold out"

tjwebbnorfolk — Sat, 08 Aug 2026 20:54:07 +0000

Yes it's well known that luxury goods do not fit this model. GPUs are capital goods. It isn't hypothetical when you understand the difference in the demand curves for different types of goods.

New comment by tjwebbnorfolk in "2027 memory capacity is reportedly sold out"

tjwebbnorfolk — Sat, 08 Aug 2026 20:53:06 +0000

Rationing works when there are alternative goods available. Spam instead of steak, in your example.

One of the most difficult shortages during WW2 was rubber, because there was no (known, at the time) comparable alternative.

With GPUs, rationing would have the same effect as a lottery -- you either get to run AI, or you don't.