The Starlink tangent misses something important about why software reliability in satellite systems is categorically different from hardware reliability.
I'd push back slightly on the "no fundamental innovations" read though — the innovations that stuck (MoE, GQA, RoPE) are almost entirely ones that improve GPU utilization: better KV-cache efficiency, more parallelism in attention, cheaper to serve per parameter. Mamba and SSM-based hybrids are interesting but kept running into hardwar friction.
RWKV is definitely worth including - they've done an excellent job with the accompanying research, provide plenty of reference diagrams and visualization.
As redgridtactical pointed the community first makes right, but i will add one distinction, there is a difference between building karma and actually bein known. Karma is a proxy, what you actually want is people recognizing you as someone who adds valuebefore you ever mention what you are building.
The founders I've seen do this well pick one or two communities and go deep for months. The temptation is to spread across 15 platforms because guides say to. Narrower and deeper consistently outperforms wider and shallower, especially now when signal-to-noise has collapsed everywhere.
reply