Replit CEO Amjad Masad: general models should JIT-train their own smaller replacements
amasad · x · 2026-10-04
In a clip shared by a16z, Replit CEO Amjad Masad argues the hot topic shouldn't just be recursive self-improvement, but "models training their replacements."
- He draws an analogy to JIT compilers: just as an interpreter spots optimization opportunities while executing dynamic code and emits optimized machine code on the fly, large general models (he cites Operator and Astra) could detect that a use case is narrow and train a small domain-specific model in real time to take over.
- The pitch is "just-in-time" distillation: general models act as discoverers and trainers, crystallizing repetitive narrow workloads into cheaper, faster specialist models.
- He closes with "Who's building this?", framing it as an open startup/research opportunity.
More from AGI Musings
- AI-powered forensic metascience: auditing 7 million papers, 500K files scanned, 100+ issues found — gleech · 2026-11-20
- Long Now Labs' first awardees include a thermodynamic clock for LLMs, $25k each — begusgasper · 2026-11-20
- Long Now launches Epistemic Cycles Lab led by Dr. Nina Beguš to map information ruptures — begusgasper · 2026-11-20
- Talia Ringer starts an email thread on open-model advocacy in a 'post-Jensen' world — dinabass · 2026-11-20
- Andriy Burkov's True Positive Weekly #186: what to know when agents do the work — burkov · 2026-11-20
- Brand drift in the agent era: give agents the reasoning behind each rule — round · 2026-11-20