June 27, 2026
Fast lane, comment pileup
Faster KNN search in Manticore: 2-pass HNSW, batched distances, and AVX-512
Manticore says searches got way faster, but the comments instantly asked: cool... where’s auth
TLDR: Manticore made its search engine noticeably faster without requiring users to change anything, which is a meaningful win for apps that rely on quick results. But commenters immediately shifted the spotlight to a more basic question — security and missing essentials — while one joker turned the whole thing into a traffic gag.
Manticore just dropped a classic "good news, no extra work required" update: its search system can now find similar items faster — in some cases by nearly 30% — thanks to behind-the-scenes tuning. Translation for normal people: if you use Manticore to power AI-style search, recommendation, or matching features, the engine should feel snappier without forcing you to rebuild anything or change settings. That alone should have been a clean victory lap.
But the real action was in the comments, where the community did what communities do best: immediately ignore the speed brag and drag the conversation somewhere messier. One of the strongest reactions was basically, "Nice speed boost, but did you finally add auth?" That turns the whole post into a mini product accountability moment. The vibe was clear: some readers are less impressed by performance gains when a major practical blocker still feels unresolved.
And then came the curveball joke: "Where are the oncoming lanes?" — a gloriously absurd dad-joke drive-by playing on the word "Manticore" like it was a road traffic update. So while the company was talking about surgical engineering wins, the crowd split into two camps: the serious users demanding missing basics, and the chaos goblins showing up for wordplay. In other words, a very online launch: impressive upgrade, instantly hijacked by feature requests and comedy.
Key Points
- •The article reports three HNSW search optimizations in Manticore: 2-pass HNSW, batched distance calculations, and AVX-512.
- •Manticore’s KNN search is built on top of the open-source hnswlib implementation.
- •The post says KNN throughput improved by up to 29% at high k.
- •The article reports more than 20% gains under concurrent load.
- •The improvements are presented as internal changes requiring no API changes, index rebuilds, or configuration updates.