Dev known for tiny-LLM work recounts dropping out, claims near-SOTA models under 8GB RAM
gnukeith · x · 2026-11-20
Security dev gnukeith shares a long personal post: he studied CS and ethical hacking, idolized Kevin Mitnick, then dropped out with 1-2 semesters left out of boredom during the GPT-3.5 era, after a professor said future devs would just be prompting AI.
Key points:
- Worked at Brave and later Obscura; now calls himself an "architect + researcher" rather than a programmer
- Increasingly obsessed with using AI for exploration and multi-angle problem solving, not just coding
- Claims to make LLMs "smaller yet smart", achieving what people call "SOTA" on under 8GB of RAM — but says it takes lots of time and compute, with no details given
- Posts an ASML machine photo: "to create something small you need a big solution"
KOL personal narrative plus an unverified claim.
More from Companies & People
- Anthropic reportedly lets employees blacklist anyone from being hired — dbasch · 2026-11-20
- Bloomberg: Apple building its own autonomous AI agent, launch possibly late next year — kimmonismus · 2026-11-20
- Proposal: Grok bots should get YouTube-style usage revenue share, Nat Friedman backs it — nateliason · 2026-11-20
- patio11 skewers governments doubting SF's AI buildout: 'You don't have enough money.' 'False.' — NathanpmYoung · 2026-11-20
- patio11: NYC sees AI labs as giant SaaS shops, SF says you don't get what they're building — NathanpmYoung · 2026-11-20
- Frontier labs may be the first true 'corporate persons': Claude can onboard anyone, no one is irreplaceable — teortaxesTex · 2026-11-20