The AI Fix Did I just start the singularity on my laptop?
13 snips
Mar 17, 2026 Corey Noles, a hands-on AI practitioner who builds small models, walks through using Codex to train a tiny language model on his laptop. He recounts hitting security roadblocks, details model architecture and training logistics, and shares why smarter models can make messier, incoherent mistakes. They also touch on industry news like layoffs, lawsuits, big funding, and a benchmark testing model nonsense.
AI Snips
Chapters
Transcript
Episode notes
December Shift Made AI Suddenly Practical
- Rapid capability jumps (a 'ChatGPT moment') in recent months made AI suddenly viable for many real-world tasks.
- Corey and Mark point to developments around Gemini 2.5 Pro and GPT-5 as catalysts for huge leaps in usefulness.
Block Fires 4,000 Citing AI Productivity
- Jack Dorsey announced Block would cut 4,000 jobs attributing it to AI-driven productivity gains.
- He claimed order-of-magnitude model improvements prompted restructuring toward smaller, flatter teams within a year.
Reddit Traps Perplexity With A Honeypot Post
- Reddit alleges Perplexity used disguised industrial-scale scraping to ingest Reddit content despite licensing deals being available.
- Reddit even planted a Google-only test post that quickly appeared in Perplexity results as evidence.