Hacker News: trenchgun

New comment by trenchgun in "Open source AI must win"

trenchgun — Sat, 13 Jun 2026 05:26:52 +0000

>But when people think of decentralized training, they don’t first think of gigantic datacenters, owned by the same company, training models across large distances. Instead, they imagine thousands of small datacenters, or individual consumers, pooling their spare compute over the internet to orchestrate a training run larger than any single actor could manage alone. Many companies are pursuing this vision: Pluralis Research, Prime Intellect and Nous Research have already successfully decentrally trained models at scale. But in practice, training decentrally over the internet has lagged far behind more centralized training. Even their largest models (Pluralis’ 8B Protocol Model, Prime Intellect’s INTELLECT-1, and Nous’ Consilience 40B) have been trained with 1,000x less compute than today’s frontier models (such as xAI’s Grok 4). https://epoch.ai/gradient-updates/how-far-can-decentralized-...

New comment by trenchgun in "xAI is looking more like a datacentre REIT than a frontier lab"

trenchgun — Mon, 08 Jun 2026 19:30:47 +0000

Yes, and then you need to have the datacenter. Do you get a permit? How long does it take to build it?

New comment by trenchgun in "Failing grades soar with AI usage, dwindling math skills in Berkeley CS classes"

trenchgun — Thu, 04 Jun 2026 07:05:21 +0000

This pattern reads heavily LLM in style, but also... that is spot on.

New comment by trenchgun in "Agentic Mfw"

trenchgun — Wed, 03 Jun 2026 06:42:39 +0000

So... you did not read it, or did not understand what you read.

New comment by trenchgun in "What are locusts and what happened to them?"

trenchgun — Sun, 31 May 2026 06:26:51 +0000

Where?

New comment by trenchgun in "Incident with Pull Requests, Issues, Git Operations and API Requests"

trenchgun — Wed, 27 May 2026 15:49:51 +0000

But everything else still works fine, right?

New comment by trenchgun in "When everyone has AI and the company still learns nothing"

trenchgun — Tue, 05 May 2026 17:29:54 +0000

Article is mostly LLM written slop.

New comment by trenchgun in "Do I belong in tech anymore?"

trenchgun — Sun, 26 Apr 2026 05:45:25 +0000

CEO observes performative work, and his inference will be that means more people need to be fired. Let only the AI native, customer obsessed 10x engineers(/ AI swarm managers) remain.

New comment by trenchgun in "Category Theory Illustrated – Orders"

trenchgun — Sat, 18 Apr 2026 17:25:10 +0000

This is LLM written comment

New comment by trenchgun in "Category Theory Illustrated – Orders"

trenchgun — Sat, 18 Apr 2026 17:19:07 +0000

How is this useful?

New comment by trenchgun in "Copilot is 'for entertainment purposes only', per Microsoft's terms of use"

trenchgun — Mon, 06 Apr 2026 07:44:34 +0000

Which of the 78 copilots? https://teybannerman.github.io/strategy/2026/03/31/how-many-...

New comment by trenchgun in "Caveman: Why use many token when few token do trick"

trenchgun — Sun, 05 Apr 2026 14:31:19 +0000

You’re a literature cycle behind. ‘Middle-layer shared representations exist’ is the observed phenomenon; ‘why exactly they form’ is the theory.

You are also confusing ‘mechanistic explanation still incomplete’ with ‘empirical phenomenon unestablished.’ Those are not the same thing.

PS. Em dash? So you are some LLM bot trying to bait mine HN for reasoning traces? :D

New comment by trenchgun in "Claude Code runs Git reset –hard origin/main against project repo every 10 mins"

trenchgun — Mon, 30 Mar 2026 06:48:39 +0000

Yes but this is probabilistic. Skill, documentation etc help by giving it the information it needs. You are then in the more correct probability distribution. Fine for docs, tips etc, but not good enough for mandatory things.

New comment by trenchgun in "The first 40 months of the AI era"

trenchgun — Sun, 29 Mar 2026 08:26:14 +0000

Yes this + also APIs! Due to LLMs needing APIs suddenly things that were strictly behind GUI can easily opened up for programmatic use. Double edged sword, due to infosec etc concerns, but sure is nice for personal productivity automation.

New comment by trenchgun in "Show HN: Cq – Stack Overflow for AI coding agents"

trenchgun — Tue, 24 Mar 2026 15:54:21 +0000

Do you run security review by agents over this?

New comment by trenchgun in "Regex Blaster"

trenchgun — Sun, 22 Mar 2026 05:30:07 +0000

Claude default frontend skills suck :D

New comment by trenchgun in "Wayland set the Linux Desktop back by 10 years?"

trenchgun — Fri, 20 Mar 2026 05:30:53 +0000

Why is it better than native x11?

New comment by trenchgun in "Leanstral: Open-source agent for trustworthy coding and formal proof engineering"

trenchgun — Tue, 17 Mar 2026 06:08:57 +0000

>Such is very difficult at the moment

What do you mean? It's a nice and simple language. Way easier to get started than OCaml or Haskell for example. And LLMs write programs in Lean4 with ease as well. Only issue is that there are not as many libraries (for software, for math proofs there is plenty).

But for example I worked with Claude Code and implemented a shell + most of unix coreutils in like a couple of hours. Claude did some simple proofs as well, but that part is obvs harder. But when the program is already in Lean4, you can start moving up the verification ladder up piece by piece.

New comment by trenchgun in "1M context is now generally available for Opus 4.6 and Sonnet 4.6"

trenchgun — Sat, 14 Mar 2026 10:13:54 +0000

LLMsdo great with Rust though

New comment by trenchgun in "Show HN: LocalGPT – A local-first AI assistant in Rust with persistent memory"

trenchgun — Sun, 08 Feb 2026 07:38:52 +0000

You could have a multi agent harness that constraints each agent role with only the needed capabilities. If the agent reads untrusted input, it can only run read only tools and communicate to to use. Or maybe have all the code running goin on a sandbox, and then if needed, user can make the important decision of effecting the real world.