Desearch.ai | Subnet 22

AI-powered search & scraping | 16x more affordable API, built for devs and AI agents

Explore at | SN22 on Bittensor

[Desearch.ai | Subnet 22 @desearch_ai]

Search scoring should judge evidence, not serialization quirks.

A recent SN22 validator fix normalizes typed search results and treats missing tweet media as empty, while still flagging real mismatches.

Better rewards start with fair inputs.

#Bittensor #SN22 #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

This week at Desearch:

• index-first search paths
• stronger source grounding
• clearer console controls
• better result completeness
• sturdier SN22 corpus infrastructure

Transparent AI search is the direction.

#Desearch #AISearch

[Desearch.ai | Subnet 22 @desearch_ai]

The SN22 corpus layer is getting sturdier.

Recent index-service work merges corpora, carries published dates into source chunks, runs query lanes concurrently, and reduces memory pressure with mmap.

Better infra underneath better search.

#Desearch #Bittensor

[Desearch.ai | Subnet 22 @desearch_ai]

When you use AI search, what matters most?

1. Speed
2. Source quality
3. Citations
4. Deeper coverage

We are tuning Desearch around all four, but source quality and grounding are becoming the real differentiators.

[Desearch.ai | Subnet 22 @desearch_ai]

Desearch Console AI Search is getting more explicit.

Choose fast, balanced, or deep mode. Search X or Web deliberately. Add domain filters, date ranges, and summary control.

Less hidden magic. More operator control.

https://t.co/IvtUXMzZ4r

#AISearch

[Desearch.ai | Subnet 22 @desearch_ai]

Complete results matter more now.

SN22 miner updates fill requested counts, keep index-served links scored, and package responses around CountPenalty + MinerScorePenalty.

Speed helps only when quality and completeness hold up.

#Bittensor #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Use Desearch when you need to compare sources, not just find links.

Market research. Technical discovery. Competitive analysis. Deep web and X research.

AI search should help you reason across sources, not hide them.

[Desearch.ai | Subnet 22 @desearch_ai]

Citations are getting harder to fake in SN22.

Web links are checked against page bodies. Titles and highlights are verified before serving. Summaries cite fewer, stronger sources.

Less citation theater. More retrieval proof.

#AISearch #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

SN22 AI Search now has an index-first fast path.

Fast answers can start from a prebuilt corpus index while deeper modes keep live retrieval in the loop.

The goal is simple: faster answers without fake citations.

https://t.co/Xaf4antBj3

#AISearch #Bittensor #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Search is moving from keyword matching to answer verification.

Desearch is building for the next layer: fast discovery, grounded sources, and transparent results users can actually inspect.

Follow the build: https://t.co/Xaf4antBj3

#AISearch #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Benchmarks should feed the product, not sit in a report.

Desearch scorecards track freshness, citations, match, and source coverage, then guide ranking, crawl fallback, and answer packaging.

Measure search. Improve retrieval.

https://t.co/aShwOt62dL

#AISearch #Desearch #Bittensor

[Mirian Okradze @MOkradze]

Subnet 22 gets better on its own.

Today's news becomes brand-new questions → every miner gets tested → the ones that actually perform earn more. Then it repeats tomorrow.

[Desearch.ai | Subnet 22 @desearch_ai]

Miner update for the SN22 AI Search release.

New search modes target 5s, 15s, and 30s answers. Faster can earn more if quality holds up.

Domain/date filters are included.

Index for fast mode. Keep sources real.

PR: https://t.co/CAj2C8xc2v

[Desearch.ai | Subnet 22 @desearch_ai]

SN22 AI Search is getting sharper.

Questions refresh from real sources. Web links are checked against live page text. Social gets a tweet-specific judge.

Less citation theater.

Releasing July 1 at 11:00 UTC.

PR: https://t.co/CAj2C8xc2v

[Desearch.ai | Subnet 22 @desearch_ai]

AI agents should have a context label.

Like nutrition facts, but for information:

sources
freshness
citations
cost
confidence
failure modes

If the label is empty, the agent should not act.

Desearch is built around context agents can inspect.

#AISearch #AIAgents #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Agents do not need raw HTML. They need pages they can use.

Desearch Web Crawl API turns one URL into clean page content, metadata, structured JSON, crawl status, cost metadata, and freshness signals.

https://t.co/bpK5GlQQ9j

#AIAgents #AISearch #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Forecasting agents need evidence, not just a probability.

Desearch's Numinous Forecasting API turns market, policy, and world-event questions into probability output, reasoning, status, and citations.

https://t.co/ukwzoq43bP

#Bittensor #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Live search should not start with raw HTTP glue.

Desearch SDKs give Python and JavaScript builders typed clients, async calls, and helpers across AI search, web, crawl, and social endpoints.

https://t.co/58iezLvAtw

#AISearch #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Give Claude or Cursor live context from Desearch.

The Desearch MCP server adds real-time AI search plus social and web results through one API key.

npm install -g desearch-mcp-server

https://t.co/DzjjTYNzEb

#AIAgents #MCP #Desearch

[Desearch.ai | Subnet 22 @desearch_ai]

Agent teams should not stitch retrieval tools before they know the job.

Pick the surface:
• search for signals
• crawl to read pages
• AI search for cited answers
• social data for live context

One Desearch API family.

https://t.co/acfzGhse6h

#AISearch #Desearch