Dev: Gemini 4 Neon rewrote my training kernels and supervised runs; rumors point to bigger models
xennygrimmato_ · x · 2026-11-23
A developer reports Gemini 4 Neon completely rewrote kernels and supervised/adjusted training runs in his pipeline, feeling beyond other frontier models. Quoted unverified rumors: Google engineers say Neon is a small model not built to win everything, with much larger models (reportedly 12T-parameter pretraining base) still coming.
More from coding & agent
- Researcher argues file systems are the wrong abstraction for LLM context layers — sh_reya · 2026-11-23
- Perplexity CEO touts zero switching cost across models with persistent sessions and AI coworkers — AravSrinivas · 2026-11-23
- Agent-built sandbox platform hits ~1.2k creates/sec, built in a week by 1.5 engineers and 300+ agents — _arohan_ · 2026-11-23
- Claude Code + video MCP: 4.5 hours of GTA 6 footage turned into 8 clips in seconds — aryanXmahajan · 2026-11-23
- Anthropic details always-on agent architecture: event inbox, timers, file memory, permissions, outbox — Arindam_1729 · 2026-11-23
- Agent OS vision: MulticaAI dashboard shows Lead, Backend and QA agents collaborating on a work board — jiayuan_jy · 2026-11-23