One unified API for AI retrieval over your database
Intelligence is only as good as its context
What WonderSearch launched
Today we’re introducing WonderSearch (YC F26) Search so good it feels like magic. No embeddings required. @ycombinator @fdotinc @wondersearchai https://t.co/DoAyAxsN8l
How TLVC rated the launch
TLVC Rating
How the launch performed
- 8.3K
- views on X
- #953
- of 1021 launches we track
- 0.0x
- the 231K median launch
What this launch shows
| Group | Views |
|---|---|
| WonderSearch, this launch | 8,283 |
| Similar-sized accounts, median across 191 launches | 198,881 |
| AI search & research launches, median across 12 launches | 242,564.5 |
| Every launch, median across 1021 launches | 231,399 |
WonderSearch reached 8.3K views, 0.0 times the directory median of 231K, ranking 953 of 1021 launches we track. Among AI search & research launches it ranks 11 of 12, 0.0 times that group's median. The post came from an account with 4.2K followers and reached 2.0 times that many people, ranking 175 of 191 launches from accounts of a similar size.
Replies ran at 21.7 per 10,000 views, well above the 3.3 median. Conversation in the first hours is what the ranking model rewards with more reach. Bookmarks came in at 76.1 per 10,000 views against a median of 13.4. There was one bookmark for every 2.5 likes.
It went out on a Saturday.
Same company
More from WonderSearch
More launches in its category
ObjectionTHE AI TRIBUNAL OF TRUTH16M views
Perplexity: Personal ComputerAI-powered answer engine14M views- Perplexity: Portable ComputerPersonal Computer is here1.2M views
ExaAI: Exa SnapshotThe world's data for AI agents730K views
Glean: Glean TauThe Work AI platform connected to all your data535K views
About the launch
WonderSearch lets teams and developers search documents by meaning without embedding the corpus first or keeping a vector index. Users upload PDFs, text, or markdown, ask in ordinary language, and receive relevant passages with source references. Search compute is charged per query, so idle collections do not keep servers running. It is built for people working through research and company knowledge, and for teams that want the same search inside a product, help center, or AI agent. Access is through MCP, a REST API, a CLI, and a Python SDK, so an agent can query a collection without a separate ingestion pipeline. The launch targets the usual first step of embedding every file before anyone has asked a question, a cost that grows when document sets are large or keep changing. WonderSearch comes from Polygres, a Y Combinator F26 company also backed by Founders, Inc. Co-founders Dalton Prescott Ng, Damien Lim, and Dale Everett Ng are the team behind it.
Tags
Questions
Frequently asked questions
The WonderSearch launch post reached 8.3K views on X, with 158 likes, 12 reposts, 18 comments, 63 bookmarks.
WonderSearch launched on X on October 17, 2026 with the message "Intelligence is only as good as its context".
WonderSearch lets teams and developers search documents by meaning without embedding the corpus first or keeping a vector index.
No, TLVC did not work on the WonderSearch launch. TLVC has done similar launches such as Tauvus, Wispr Flow, and Airtable - Hyperagent.
TLVC rated the WonderSearch launch 3.20 out of 5: hook 4/5, editing / creativity 3/5, copy 4/5, sentiment of launch 4/5, distribution strategy 1/5.
