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]

New blog how-to is live: How to Find Reddit Threads by Keyword With a Web Search API

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

People Search is in the Console.

Describe the people you want, for example by what they post about. Each profile comes back with the reasons it matched, so you can check the list instead of guessing.

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Pricing changes should be diffs, not memories.

This open Desearch workflow saves plan names, prices and billing notes, then compares the next snapshot.

If a page shows no prices, it says so instead of guessing.

Code: https://t.co/clQqMojgNL

#Desearch #AIAgents #OpenSource

[Desearch.ai | Subnet 22 @desearch_ai]

Company Search is in the Console.

Describe the companies you want in plain words. Each result gets an AI score, a short why-this-matched note, and a check for each criterion, so you can see why a company made the list.

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Star counts show movement, not why.

Repo Buzz Tracker combines public mentions, replies and web pages into a dated evidence file. The query, dates and match type stay visible.

https://t.co/TM42UFB3jN

#Desearch #OpenSource #SocialData

[Desearch.ai | Subnet 22 @desearch_ai]

Desearch Monitors are live in the Console.

Save a keyword or hashtag once. It runs on a schedule and keeps only what is new across X, Reddit, YouTube, the web and more. Add a webhook if you want a ping when something matches.

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Deep-research reports should be rebuildable, not just readable.

This Desearch workflow saves raw responses, uses dated pages for findings, and keeps exact quotes + source dates.

Run it:
https://t.co/KpRCwKte2f

#Desearch #AISearch #OpenSource

[Desearch.ai | Subnet 22 @desearch_ai]

New blog how-to is live: How to Get the Retweeters of an X Post With an API

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Before grading an answer, ask if retrieval found the right page.

On SimpleQA Verified, Desearch put the reference page in the top 10 for 99.1% of 463 questions.

Retrieval is one layer, not answer correctness.

Method: https://t.co/Z5sfAC1BhI

#Desearch #AISearch

[Desearch.ai | Subnet 22 @desearch_ai]

What is trending where, right now?

Give this Python script a list of cities and countries. It pulls the current X trends for each, fetches the top posts behind the leading trends, and shows which trends are shared and which are local.

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Do not let the loudest complaint define the whole pattern.

Open Complaint Radar dedupes public posts, groups issues by theme and keeps source links for review.

Code:
https://t.co/d0NE4cz8EO

#Desearch #SocialData

[Desearch.ai | Subnet 22 @desearch_ai]

What do your comments actually say?

Give this Python script a brand's Instagram and TikTok handles. It sorts the comments on recent posts into questions, complaints, requests and praise, with the most-liked ones in each group.

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Desearch MCP now supports Meta Muse connector OAuth.

Muse can dynamically register with https://t.co/5ZC5YcqbRK, while tool calls still require user authorization.

Personal agents get live Desearch context without a custom auth bridge.

#Desearch #MCP #AIAgents

[Desearch.ai | Subnet 22 @desearch_ai]

What ads is that store running this week?

Give this Python script a few store names. It reads their public Facebook Ad Library records, groups copies of the same ad, tags each angle and offer, and shows what is new, stopped or unchanged.

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Company research should be bounded and resumable.

Desearch Company Search can stop at a USD budget. If coverage is still thin, continue from an earlier task without returning the same companies.

Docs: https://t.co/tsav1Twpvh

#Desearch #AIAgents

[Desearch.ai | Subnet 22 @desearch_ai]

A few API notes.

Social list endpoints now return a next-page cursor header.
Instagram, TikTok and Facebook results have new fields.
X results now include the post card.
Crawl takes a URL without https:// at the front.

Docs link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

Not every search has the same goal.

Use fast mode when latency blocks the workflow. Use balanced mode when relevance matters more than a shorter wait.

Choose per task, then measure.

https://t.co/DtLOOpjh4U

#Desearch #AISearch #AIAgents

[Desearch.ai | Subnet 22 @desearch_ai]

Who starts Black Friday first?

This small Python script takes a dated snapshot of a few brands each day: sale page banners, their latest X, Instagram and TikTok posts, and a web search for their Black Friday page.

By late November you have a timeline.

Link on LinkedIn.

[Desearch.ai | Subnet 22 @desearch_ai]

A directory list is not a distribution plan.

This Desearch example dedupes results, reads likely submit pages, and keeps evidence about free, paid and review rules.

https://t.co/UEWybNQnlO

#Desearch #AISearch #OpenSource

[Desearch.ai | Subnet 22 @desearch_ai]

"Best tools" pages send buyers straight to the brands they list. AI answers read them too.

This script takes your brand, 3 to 5 others in your category and 5 to 8 buyer searches, then flags the pages that name 2 or more of them but skip you.

Link on LinkedIn.