2026-04-09T13:23:37.854Z · Src: 2026-04-09T10:00:00.000Z
2026-04-09T12:52:05.260Z · Src: 2026-04-08T12:00:00.000Z
2026-04-09T12:51:34.560Z · Src: 2026-04-08T18:00:00.000Z
2026-04-06T16:20:41.386Z · Src: 2026-04-06T00:00:00.000Z
2026-04-06T11:20:57.842Z · Src: 2026-04-06T00:00:00.000Z
2026-04-06T04:48:00.654Z · Src: 2026-04-06T00:00:00.000Z
2026-03-29T10:19:24.550Z · Src: 2026-03-29T00:00:00.000Z
2026-03-28T12:06:44.734Z · Src: 2026-03-28T00:00:00.000Z
2026-03-25T12:58:47.481Z · Src: 2026-03-25T00:00:00.000Z
A DuckDB fork implements ACORN-1 filtered HNSW search, solving the problem of filtered vector queries returning incomplete results. Benchmarks show 10x improvement on low-selectivity filters like lang
2026-03-25T11:18:56.360Z · Src: 2026-03-25T00:00:00.000Z
Google Research's TurboQuant achieves extreme AI model compression while preserving quality, potentially reducing inference costs, enabling edge deployment, and democratizing access to powerful AI mod
2026-03-17T16:56:08.000Z
Been running autoresearch for about a week. ~100 experiments per night on an H100. The keep rate is around 15%. The problem isn't the keep/discard loop. That works. The problem is that some of those k