Hacker News: sirsinsalot

New comment by sirsinsalot in "Claude Fable 5"

sirsinsalot — Tue, 09 Jun 2026 20:25:33 +0000

I think intent, rather than target, is implied and important.

You should see the abuse my motorbike gets. Poor thing.

New comment by sirsinsalot in "Ask HN: Why hasn't there been a real competitor to Ticketmaster yet?"

sirsinsalot — Mon, 08 Jun 2026 20:08:56 +0000

They also own a lot of the venue infrastructure across the industry such as catering, tour buses, security,...

They have leverage with venues they dont own and a monopoly across industry verticals.

Sickening situation for music.

New comment by sirsinsalot in "Building from zero after addiction, prison, and a felony"

sirsinsalot — Mon, 08 Jun 2026 00:32:29 +0000

I ride. No way in hell I'd ride in the US.

New comment by sirsinsalot in "Google to pay SpaceX $920M a month for compute capacity at xAI data centers"

sirsinsalot — Sat, 06 Jun 2026 22:44:31 +0000

Not doing much to beat the accusations of circular dealing are they?

New comment by sirsinsalot in "How LLMs work"

sirsinsalot — Sat, 06 Jun 2026 15:37:39 +0000

I suspect it was considered many times, but the sheer computation scale would make it feel like obscene brute force. It feels like the right shape but too wild to think about implementing.

I think as time went on, and hardware got better, it seemed more reasonable to actually think about a viable implementation of what I think was a widespread intuition anyone in ML had that everything's context is everything.

It just seemed like a theoretical thing until hardware caught up. Maybe. Perhaps I'm applying a retrospective excuse to why it took so long.

New comment by sirsinsalot in "Moving beyond fork() + exec()"

sirsinsalot — Sat, 06 Jun 2026 15:32:31 +0000

I have a rule for myself. If I think something is silly or stupid, I assume I don't understand it. I usually find I do not understand it, and it no longer seems silly when I do understand it.

In this case too, you think it is silly because you don't understand it. Your assumptions are wrong, making it seem silly.

New comment by sirsinsalot in "GrapheneOS user reported to authorities for using GrapheneOS"

sirsinsalot — Sat, 06 Jun 2026 10:29:09 +0000

But get your car stolen in the UK and the police won't do a thing. Even if you know where it is via a tracker. Nothing. Outright refuse to take any action.

New comment by sirsinsalot in "How LLMs work"

sirsinsalot — Sat, 06 Jun 2026 10:24:27 +0000

I think the attention mechanism is so simple but so revolutionary that people forget it.

Like the best leaps in thinking, once it is made, is is immediately obvious and intuitive.

New comment by sirsinsalot in "How LLMs work"

sirsinsalot — Sat, 06 Jun 2026 10:19:56 +0000

You missed the third and most important reason to learn: fun.

Which sums up HN these days.

New comment by sirsinsalot in "When AI Builds Itself: Our progress toward recursive self-improvement"

sirsinsalot — Fri, 05 Jun 2026 07:26:00 +0000

Truly feels like witnessing the worst of capitalism and greed play out. All that compute and energy towards a narrative of reducing the need for skilled programmers. What a waste.

These people don't have our interests in mind and everyone eats it up like a blessing from a god or something. It's surreal.

New comment by sirsinsalot in "Uber's $1,500/month AI limit is a useful signal for AI tool pricing"

sirsinsalot — Wed, 03 Jun 2026 23:51:36 +0000

How dare you mention evidence! This isn't engineering you know!

New comment by sirsinsalot in "Uber's $1,500/month AI limit is a useful signal for AI tool pricing"

sirsinsalot — Wed, 03 Jun 2026 23:44:48 +0000

Fear of loss to competitors embracing a technology creates a fear driven adoption.

Let me ask you this: is any technology worth so much break-neck adoption without first seeing clear evidence of ROI? No. The adoption is irrational.

New comment by sirsinsalot in "Alphabet announces $80B equity capital raise to expand AI infra and compute"

sirsinsalot — Tue, 02 Jun 2026 10:38:17 +0000

I think Google is doing the right thing. Using LLMs for coding is the shiny low hanging fruit but it isn't what is going to make the tech ubiquitous. That'll be finding applications of it to real data problems.

Google knows LLMs are the new UI, not the new IDE.

New comment by sirsinsalot in "Constraint Decay: The Fragility of LLM Agents in Back End Code Generation"

sirsinsalot — Mon, 25 May 2026 14:06:14 +0000

My point was that if training data + encoding/training = model with emergent behaviour

The emergent behaviour is in the training data and/or encoding/training.

So while I agree it is emergent from the complexity, it isn't some unknown mechanism. Just complexity at scale.

New comment by sirsinsalot in "Constraint Decay: The Fragility of LLM Agents in Back End Code Generation"

sirsinsalot — Mon, 25 May 2026 12:03:25 +0000

Not to be pedantic but the emergent properties are in the training set, and thus the model and algorithm. There's no magic coming from the universe.

What makes the behavior emergent is that it can't be predicted at training time.

The emergent and unpredictable output is the result of massive vector complexity being encoded.

New comment by sirsinsalot in "Texas woman arrested for Facebook post about town water quality"

sirsinsalot — Sat, 23 May 2026 20:31:30 +0000

Perhaps in the first order, but when premiums go up and go up across all policies due to the acceptability of litigation... Everyone pays eventually.

Its a bit like saying driving dangerously is OK because you have insurance. Until everyone drives dangerously and insurance is sky high for all.

That said, they should be sued.

New comment by sirsinsalot in "Texas woman arrested for Facebook post about town water quality"

sirsinsalot — Sat, 23 May 2026 20:27:20 +0000

Because the US is a third world country cosplaying as a developed nation. Much like their president is a corrupt and morally bankrupt fool cosplaying as a politician.

It doesn't matter in the US. Just pretend.

New comment by sirsinsalot in "Deno 2.8"

sirsinsalot — Sat, 23 May 2026 00:18:58 +0000

Or at least some kind of cursor based text buffer tree that wasm can bind to the dom/shadow tree, rather than a fuller API

New comment by sirsinsalot in "Deno 2.8"

sirsinsalot — Sat, 23 May 2026 00:17:19 +0000

As I understand it you can, in fact node 23 has a type stripping feature. I could be wrong.

New comment by sirsinsalot in "Deno 2.8"

sirsinsalot — Sat, 23 May 2026 00:16:15 +0000

I mean, all typescript does is serve no purpose except for an LSP and documenting type intention really. It isn't much different to Python type annotation.