# 2026 W02 - 01-11 - Pro or Cons LLM, that's not the question

First time here ? Check the introduction [here]({% post_url hn_digest/2026-01-01-Init %})


# Articles 

## Model Training

- 327 pt  [[Link](<https://github.com/samuel-vitorino/sopro>) / [HN](<https://news.ycombinator.com/item?id=46546113>)] Sopro TTS: A 169M model with zero-shot voice cloning that runs on the CPU 
- 321 pt  [[Link](<https://byteshape.com/blogs/Qwen3-30B-A3B-Instruct-2507/>) / [HN](<https://news.ycombinator.com/item?id=46518573>)] A 30B Qwen model walks into a Raspberry Pi and runs in real time 

## Coding Agents

- 784 pt  [[Link](<https://burkeholland.github.io/posts/opus-4-5-change-everything/>) / [HN](<https://news.ycombinator.com/item?id=46515696>)] Opus 4.5 is not the normal AI agent experience that I have had thus far 
- 492 pt  [[Link](<https://github.com/anomalyco/opencode/issues/7410>) / [HN](<https://news.ycombinator.com/item?id=46549823>)] Anthropic blocks third-party use of Claude Code subscriptions 
- 397 pt  [[Link](<https://spectrum.ieee.org/ai-coding-degrades>) / [HN](<https://news.ycombinator.com/item?id=46542036>)] AI coding assistants are getting worse? 
- 136 pt  [[Link](<https://jakobemmerling.de/posts/fuse-is-all-you-need/>) / [HN](<https://news.ycombinator.com/item?id=46580136>)] FUSE is All You Need – Giving agents access to anything via filesystems 
- 133 pt  [[Link](<https://github.com/richhickson/claudecodeusage>) / [HN](<https://news.ycombinator.com/item?id=46544524>)] Show HN: macOS menu bar app to track Claude usage in real time 
- 132 pt  [[Link](<https://github.com/anthropics/claude-code/issues/16673>) / [HN](<https://news.ycombinator.com/item?id=46532075>)] Claude Code CLI was broken 
- 103 pt  [[Link](<https://www.lesswrong.com/posts/u6Lacc7wx4yYkBQ3r/insights-into-claude-opus-4-5-from-pokemon>) / [HN](<https://news.ycombinator.com/item?id=46515625>)] Insights into Claude Opus 4.5 from Pokémon 

## Politics / Strategy

- 709 pt  [[Link](<https://twitter.com/OfficialLoganK/status/2009339263251566902>) / [HN](<https://news.ycombinator.com/item?id=46545077>)] Google AI Studio is now sponsoring Tailwind CSS 
- 402 pt  [[Link](<https://openai.com/index/introducing-chatgpt-health/>) / [HN](<https://news.ycombinator.com/item?id=46531280>)] ChatGPT Health 

## Personal blogs / opinions

- 1032 pt  [[Link](<https://antirez.com/news/158>) / [HN](<https://news.ycombinator.com/item?id=46574276>)] Don't fall into the anti-AI hype 
- 455 pt  [[Link](<https://trails.pieterma.es/>) / [HN](<https://news.ycombinator.com/item?id=46567400>)] Show HN: I used Claude Code to discover connections between 100 books 
- 315 pt  [[Link](<https://dri.es/ai-is-a-business-model-stress-test>) / [HN](<https://news.ycombinator.com/item?id=46567392>)] AI is a business model stress test 
- 215 pt  [[Link](<https://www.wsj.com/tech/ai/google-ai-openai-gemini-chatgpt-b766e160>) / [HN](<https://news.ycombinator.com/item?id=46528389>)] How Google got its groove back and edged ahead of OpenAI 
- 194 pt  [[Link](<https://arpitbhayani.me/blogs/qkv-matrices/>) / [HN](<https://news.ycombinator.com/item?id=46523887>)] The Q, K, V Matrices 
- 185 pt  [[Link](<https://calnewport.com/why-didnt-ai-join-the-workforce-in-2025/>) / [HN](<https://news.ycombinator.com/item?id=46505735>)] Why didn't AI “join the workforce” in 2025? 
- 165 pt  [[Link](<https://consciousdigital.org/chatgpt-health-is-a-marketplace-guess-who-is-the-product/>) / [HN](<https://news.ycombinator.com/item?id=46541533>)] ChatGPT Health is a marketplace, guess who is the product? 
- 144 pt  [[Link](<https://matthewrocklin.com/ai-zealotry/>) / [HN](<https://news.ycombinator.com/item?id=46557057>)] My article on why AI is great (or terrible) or how to use it 
- 143 pt  [[Link](<https://llmholdem.com/>) / [HN](<https://news.ycombinator.com/item?id=46569061>)] Show HN: Play poker with LLMs, or watch them play against each other 
- 118 pt  [[Link](<https://embd.cc/llm-problems-observed-in-humans>) / [HN](<https://news.ycombinator.com/item?id=46527581>)] LLM Problems Observed in Humans 
- 109 pt  [[Link](<https://idiallo.com/blog/all-ai-videos-are-harmful>) / [HN](<https://news.ycombinator.com/item?id=46498651>)] All AI Videos Are Harmful (2025) 
- 114 pt  [[Link](<https://sakana.ai/drq/>) / [HN](<https://news.ycombinator.com/item?id=46542761>)] Digital Red Queen: Adversarial Program Evolution in Core War with LLMs 
- 89 pt  [[Link](<https://arxiv.org/abs/2512.09882>) / [HN](<https://news.ycombinator.com/item?id=46518996>)] Comparing AI agents to cybersecurity professionals in real-world pen testing 
- 59 pt  [[Link](<https://www.daily.co/blog/building-voice-agents-with-nvidia-open-models/>) / [HN](<https://news.ycombinator.com/item?id=46528045>)] Building voice agents with Nvidia open models 


## News

- 586 pt  [[Link](<https://mathstodon.xyz/@tao/115855840223258103>) / [HN](<https://news.ycombinator.com/item?id=46560445>)] “Erdos problem #728 was solved more or less autonomously by AI” 
- 116 pt  [[Link](<https://www.emjreviews.com/radiology/news/ai-misses-nearly-one-third-of-breast-cancers-study-finds/>) / [HN](<https://news.ycombinator.com/item?id=46537983>)] AI misses nearly one-third of breast cancers, study finds 
- 71 pt  [[Link](<https://arstechnica.com/tech-policy/2025/12/openai-refuses-to-say-where-chatgpt-logs-go-when-users-die/>) / [HN](<https://news.ycombinator.com/item?id=46499983>)] Murder-suicide case shows OpenAI selectively hides data after users die 
- 66 pt  [[Link](<https://www.promptarmor.com/resources/ibm-ai-(-bob-)-downloads-and-executes-malware>) / [HN](<https://news.ycombinator.com/item?id=46544454>)] IBM AI ('Bob') Downloads and Executes Malware 

## Unclassified

- 179 pt  [[Link](<https://www.promptarmor.com/resources/notion-ai-unpatched-data-exfiltration>) / [HN](<https://news.ycombinator.com/item?id=46531565>)] Notion AI: Unpatched data exfiltration 
- 67 pt  [[Link](<https://github.com/marcoaapfortes/Mantic.sh>) / [HN](<https://news.ycombinator.com/item?id=46512182>)] Show HN: Mantic.sh – A structural code search engine for AI agents