
What happens when an LLM hallucinates the Hacker News front page ten years into the future and then the rest of the internet tries to debug it? In today’s HackerNews.fm (Dec 9, 2025), we dig into a wild “Show HN” experiment with Gemini Pro 3, a brand‑new Pebble device pitched as “external memory for your brain,” and Mistral’s latest shot at autonomous coding with Devstral 2 and the Vibe CLI. Then we zoom out: from a bold new theory about a universal “weight subspace” for neural networks, to training a home‑grown LLM on a single RTX 3090, to a mind‑bending argument that time itself might not exist the way we think it does.
Stories covered in this episode:
1. 01:04 Show HN: Gemini Pro 3 Hallucinates the HN Front Page 10 Years from Today
• Rank: 1 • Points: 2299 • Comments: 722
• Article: https://dosaygo-studio.github.io/hn-front-page-2035/news
• HN Discussion: https://news.ycombinator.com/item?id=46205632
2. 04:03 New Pebble Device
• Rank: 1 • Points: 451 • Comments: 434
• Article: https://repebble.com/blog/meet-pebble-index-01-external-memory-for-your-brain
• HN Discussion: https://news.ycombinator.com/item?id=46205661
3. 07:48 Mistral Releases Devstral 2 (72.2% SWE-Bench Verified) and Vibe CLI
• Rank: 1 • Points: 549 • Comments: 272
• Article: https://mistral.ai/news/devstral-2-vibe-cli
• HN Discussion: https://news.ycombinator.com/item?id=46205437
4. 11:03 The Universal Weight Subspace Hypothesis
• Rank: 1 • Points: 346 • Comments: 122
• Article: https://arxiv.org/abs/2512.05117
• HN Discussion: https://news.ycombinator.com/item?id=46199623
5. 14:37 LLM from scratch, part 28 – training a base model from scratch on an RTX 3090
• Rank: 1 • Points: 497 • Comments: 102
• Article: https://www.gilesthomas.com/2025/12/llm-from-scratch-28-training-a-base-model-from-scratch
• HN Discussion: https://news.ycombinator.com/item?id=46124425
6. 18:39 Time might not exist – and we're starting to understand why
• Rank: 1 • Points: 70 • Comments: 72
• Article: https://www.sciencefocus.com/science/the-closer-we-look-at-time-the-stranger-it-gets
• HN Discussion: https://news.ycombinator.com/item?id=46201750