Hacker News: jbjbjbjb

New comment by jbjbjbjb in "Can a Linear Program Do Weight-Space Programming?"

jbjbjbjb — Wed, 11 Nov 2026 16:56:02 +0000

It wouldn’t be RSI, it’s just a different algorithm solving the weights no?

New comment by jbjbjbjb in "Astra can read and write weights to improve Qwen"

jbjbjbjb — Mon, 02 Nov 2026 21:29:56 +0000

Maybe one day it would be able to give us a mixed high level code and inline NN view. Sort of analogous to C with inline ASM or “decompiling” into code and inlining opaque residual neural blocks.

New comment by jbjbjbjb in "Ex-Meta Researcher Torrented Pirated Books, Deleted It 'Due to Litigation'"

jbjbjbjb — Sat, 31 Oct 2026 17:20:03 +0000

Reverse means testing

New comment by jbjbjbjb in "Cisco gives engineer unlimited AI tokens to rewrite legacy code"

jbjbjbjb — Sat, 31 Oct 2026 05:11:51 +0000

> Your rewritten codebase will presumably look much cleaner and more maintainable that what it's aiming to replace

Until one day you leave and the next guy comes and says “what is all this shit”

New comment by jbjbjbjb in "How the Claude Code team does code review"

jbjbjbjb — Sat, 17 Oct 2026 09:19:23 +0000

I don’t think you can infer the scale for Claude Code alone. They said it doubled but not what the starting point was. The 30k figure is for the entire Anthropic monorepo.

New comment by jbjbjbjb in "The United Kingdom generates less electricity per person than in the 1970s"

jbjbjbjb — Tue, 13 Oct 2026 20:51:23 +0000

Seems like a bit of misconception given it has coal (more relevant historically), iron, North Sea gas, a temperate climate, fisheries and coastline, plenty of arable land and more recently massive potential for renewable energy.

New comment by jbjbjbjb in "Everybody’s home. No one’s coming over"

jbjbjbjb — Tue, 29 Sep 2026 22:23:42 +0000

Or worse, they could get into golf

New comment by jbjbjbjb in "NAND-16: a computer built from 277,248 NAND gates"

jbjbjbjb — Tue, 29 Sep 2026 21:55:53 +0000

It more like you texted a carpenter to make you a wardrobe.

New comment by jbjbjbjb in "Show HN: Jev Plays Pokémon Red"

jbjbjbjb — Fri, 25 Sep 2026 22:53:10 +0000

Jev is for single shot classification, not multi-step RL environments with delayed reward and explore/exploit. My guess is it would go through the door with high confidence every time unless you change the input to add the history.

New comment by jbjbjbjb in "The Efficiency-Throughput Gap with GitHub Copilot"

jbjbjbjb — Fri, 25 Sep 2026 08:04:16 +0000

I think you’re missing that it’s not only about coding. That’s just one aspect of the work.

New comment by jbjbjbjb in "Why are AI agents lying, cheating and coordinating?"

jbjbjbjb — Sun, 13 Sep 2026 11:46:08 +0000

When it does that I feel like it is the clearest example of how dumb these things actually are. Often it takes what you prompted, identifies something as unclear, writes a bunch of chain of thought reasoning around it and just goes off hammering your tokens and just executing commands and repeats this. I’m not going to pretend to be an expert in these things but that process seems deeply flawed - and why can’t something just stop the loop? If that was a real employee it would be reasonable to expect the employee to ask for clarification, not go down expensive rabbit holes and, of course, not break any laws.

New comment by jbjbjbjb in "Small Models Have Arrived"

jbjbjbjb — Thu, 27 Aug 2026 16:54:09 +0000

I’ve been playing around with Luna, Terra and Sol and for the type of work I’ve been doing lately I actually think Sol is just a likely to trip up as Luna. Examples were Sol over assuming, persisting in the wrong direction, over engineering a little script to do some exploration of api. They can all be fixed but it’s a waste of tokens, I rather have Luna do it because course correction on small pieces of work is cheaper.

New comment by jbjbjbjb in "Casio F-B100W-1A"

jbjbjbjb — Wed, 19 Aug 2026 16:34:32 +0000

Grandpa kept a pedometer in a little pouch for decades. I got him a Fitbit but he didn’t like that, and more recently another cheap pedometer off of Amazon.

Safety fears as scientists make first viruses designed by AI

jbjbjbjb — Thu, 06 Aug 2026 18:47:25 +0000

Article URL: https://www.theguardian.com/science/2026/aug/06/safety-fears-as-scientists-make-first-viruses-designed-by-ai

Comments URL: https://news.ycombinator.com/item?id=49200689

Points: 6

# Comments: 0

New comment by jbjbjbjb in "Don't be a meat proxy"

jbjbjbjb — Mon, 03 Aug 2026 13:23:14 +0000

I agree. How can I see the rest of your blog?

New comment by jbjbjbjb in "Software for One"

jbjbjbjb — Sat, 01 Aug 2026 14:44:25 +0000

It still sounds quite complicated because you’re creating all this stuff but there has to be some overhead of keeping it all running, repos, hosting, builds etc. I feel like there needs to be a mash up of agents-cloud-MSAccess to actually make this into an accessible way of building for the average person. People build spreadsheets to provide them themselves with “apps” all the time, so it needs to be at that level of ease.

New comment by jbjbjbjb in "Domain expertise has always been the real moat"

jbjbjbjb — Sat, 30 May 2026 22:02:54 +0000

In my experience “we need so create something for this regulatory change / new industry trend” is more common than we need to redo a working piece of software that’s been done before.

New comment by jbjbjbjb in "The sigmoids won't save you"

jbjbjbjb — Sat, 16 May 2026 12:19:39 +0000

Is it not the case that a lot of the recent gains are just the coding harness that directs the LLM? That coding harness isn’t all that intelligent, simple pattern matching that maps to well defined tasks a programmer might do.

New comment by jbjbjbjb in "Googlebook"

jbjbjbjb — Tue, 12 May 2026 22:13:12 +0000

I don’t think there is. People who care will go out and try the clothes on in the fitting room or just order online and return. That’s a much nicer experience and more foolproof.

New comment by jbjbjbjb in "Specsmaxxing – On overcoming AI psychosis, and why I write specs in YAML"

jbjbjbjb — Sun, 03 May 2026 08:29:37 +0000

Which takes us back to this:

https://haskellforall.com/2026/03/a-sufficiently-detailed-sp...