Hacker News: armcat

New comment by armcat in "Screenshots of Old Desktop OSes"

armcat — Tue, 12 May 2026 17:31:20 +0000

Sun Solaris PPC (CDE) takes me back. I've built plenty of 3G/WCDMA telephony code on that thing. It never let me down.

ZAYA1-8B: Frontier intelligence density, trained on AMD

armcat — Thu, 07 May 2026 05:07:31 +0000

Article URL: https://www.zyphra.com/post/zaya1-8b

Comments URL: https://news.ycombinator.com/item?id=48045608

Points: 3

# Comments: 0

New comment by armcat in "Show HN: A Karpathy-style LLM wiki your agents maintain (Markdown and Git)"

armcat — Sat, 25 Apr 2026 10:06:21 +0000

Any particular reason for BM25? Why not just a table of contents or index structure (json, md, whatever) that is updated automatically and fed in context at query time? I know bag of words is great for speed but even at 1000s of documents, the index can be quite cheap and will maximise precision

Building agents that reach production systems with MCP

armcat — Thu, 23 Apr 2026 03:41:14 +0000

Article URL: https://claude.com/blog/building-agents-that-reach-production-systems-with-mcp

Comments URL: https://news.ycombinator.com/item?id=47872044

Points: 4

# Comments: 0

Andon Cafe – A cafe in Stockholm run by AI

armcat — Wed, 22 Apr 2026 18:23:31 +0000

Article URL: https://andon.cafe/

Comments URL: https://news.ycombinator.com/item?id=47867328

Points: 2

# Comments: 0

AI Agent Traps (DeepMind)

armcat — Mon, 20 Apr 2026 08:24:23 +0000

Article URL: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6372438

Comments URL: https://news.ycombinator.com/item?id=47831602

Points: 1

# Comments: 0

Ask HN: In the AI world what does "great" look like?

armcat — Fri, 17 Apr 2026 19:23:47 +0000

I had a chat with my team mate today and they brought up an interesting point. If AI is cooking our software and doing an amazing job, for us devs what does "great" look like? How do WE improve? How do we assess performance of us, i.e. "the humans"? What do we base salaries on?

Have you discussed this in your organisations? Does anyone have a plan? If your dev teams are creating scaffolding, harnesses, skills and getting CC to cook up everything, make PRs, perform reviews, find bugs, fix bugs, update architecture diagrams....what goals do you set in performance development talks?

Need some guidance.

Comments URL: https://news.ycombinator.com/item?id=47809594

Points: 3

# Comments: 1

Qwen3.6-35B-A3B draws a better pelican than Opus 4.7

armcat — Thu, 16 Apr 2026 17:50:54 +0000

Article URL: https://twitter.com/simonw/status/2044830134885306701

Comments URL: https://news.ycombinator.com/item?id=47797020

Points: 5

# Comments: 0

New comment by armcat in "Codex for almost everything"

armcat — Thu, 16 Apr 2026 17:25:10 +0000

Is it OpenAI Cowork?

ChatGPT for Excel

armcat — Wed, 15 Apr 2026 21:21:05 +0000

Article URL: https://chatgpt.com/apps/spreadsheets/

Comments URL: https://news.ycombinator.com/item?id=47785397

Points: 341

# Comments: 198

Language models transmit behavioural traits through hidden signals in data

armcat — Wed, 15 Apr 2026 19:53:25 +0000

Article URL: https://www.nature.com/articles/s41586-026-10319-8

Comments URL: https://news.ycombinator.com/item?id=47784352

Points: 4

# Comments: 2

The AI Layoff Trap

armcat — Mon, 13 Apr 2026 05:56:49 +0000

Article URL: https://arxiv.org/abs/2603.20617

Comments URL: https://news.ycombinator.com/item?id=47748123

Points: 62

# Comments: 104

New comment by armcat in "Claude for Word in Now in Beta"

armcat — Fri, 10 Apr 2026 19:49:19 +0000

As someone who's been working in legaltech space where MS Word add-in chatbot was a killer feature, this is brutal. And in their demo they are hammering on the legal case (redline chat).

Claude for Word in Now in Beta

armcat — Fri, 10 Apr 2026 19:02:11 +0000

Article URL: https://twitter.com/claudeai/status/2042670341915295865

Comments URL: https://news.ycombinator.com/item?id=47722236

Points: 6

# Comments: 3

New comment by armcat in "Components of a Coding Agent"

armcat — Sat, 04 Apr 2026 15:42:42 +0000

I still find it incredible at the power that was unleashed by surrounding an LLM with a simple state machine, and giving it access to bash

New comment by armcat in "CERN uses ultra-compact AI models on FPGAs for real-time LHC data filtering"

armcat — Sat, 28 Mar 2026 11:14:46 +0000

Not on the same extreme level, but I know that some coffee machines use a tiny CNN based model locally/embedded. There is a small super cheap camera integrated in the coffee machine, and the model does three things: (1) classifies the container type in order to select type of coffee, (2) image segmentation - to determine where the cup/hole is placed, (3) regression - to determine the volume and regulate how much coffee to pour.

Show HN: The Economics of Builder Saturation in Digital Markets

armcat — Fri, 27 Mar 2026 13:26:31 +0000

I formalised something most of us already feel but rarely say out loud: making things easier to build doesn't make things easier to succeed with.

Personal version: I've vibe-coded maybe 15 projects since the beginning of this year. Two are still alive. At work, our teams built hundreds of custom GPTs and dashboards. Handful survived. The failure mode was never "couldn't build it" - it was "nobody had the bandwidth to care."

So I wrote a paper about it. It combines Herbert Simon's attention scarcity, free-entry IO models, superstar economics, and preferential attachment into one framework. The central result: equilibrium attention per builder = k/p (entry cost over monetisation rate), independent of market size. As AI drives k towards 0, that ratio vanishes regardless of how much the market grows. Free entry absorbs everything.

Calibrated to the App Store (800K publishers, 38B downloads) the model matches observed concentration pretty well - top 1% get ~70% of downloads, quarter of apps under 100 downloads, Gini above 0.9.

The same mechanism works inside organisations (dashboard sprawl, GPT graveyards, tool fatigue) and across markets. It's the same math: finite attention, elastic production, winner-take-most.

I just got a bit fed up with the narrative I am constantly seeing online: that everyone will be a successful builder with AI; just build; forget about "everything else" that you do - if you aren't vibe coding, you aren't doing anything. And I'm saying this as an AI engineer who has built dozens of models and in the recent times many apps with Claude Code and Codex. I thought it's time to shine some mathematics on this.

Paper: https://arxiv.org/abs/2603.23685

PS. I was very inspired by Herbert Simon (Nobel Prize in Economics) and ironically enough, he is also considered one of the "founders of AI".

Comments URL: https://news.ycombinator.com/item?id=47542386

Points: 1

# Comments: 0

AI Efficiency Courses

armcat — Wed, 25 Mar 2026 19:59:17 +0000

Article URL: https://github.com/PrunaAI/ai-efficiency-courses

Comments URL: https://news.ycombinator.com/item?id=47522378

Points: 1

# Comments: 0

New comment by armcat in "Quantization from the Ground Up"

armcat — Wed, 25 Mar 2026 16:53:08 +0000

This is beautifully written and visualised, well done! The KL divergence comparisons between original and different quantisation levels is on-point. I'm not sure people realize how powerful quantisation methods are and what they've done for democratising local AI. And there are some great players out there like Unsloth and Pruna.

Design.md

armcat — Fri, 20 Mar 2026 16:21:30 +0000

Article URL: https://stitch.withgoogle.com/docs/design-md/overview/

Comments URL: https://news.ycombinator.com/item?id=47456828

Points: 2

# Comments: 0