Joined January 2025
- Tweets 538
- Following 66
- Followers 1,672
- Likes 793
209 Photos and videos
On the product side, the Console now handles link only AI Search results more clearly and shows streaming progress instead of leaving the visual panel blank while results are building.
1
2
202
On the API and subnet side, recent work tightened social result IDs, balance checks, local playground auth, synthetic scoring, duplicate penalties, and web/X relevance rules.
Small changes, but they compound into trust.
2
175
SN22 validator scoring keeps getting stricter.
The latest update diversifies synthetic AI Search questions, separates web and X relevance by tool type, includes quoted tweet content, and improves duplicate handling.
Quality should matter in rewards.
#Bittensor #SN22
1
3
13
321
Desearch is showing long-term conviction onchain.
763.7K SN22 alpha is perpetually locked under Bittensor's Conviction upgrade, about 24.1% of staked SN22 alpha.
501K by Desearch.
262.7K by external holders.
Conviction for decentralized search.
#Bittensor #SN22 #Desearch
7
25
636
IntoTAO covered Desearch clearly.
SN22 turns search into a live quality loop:
miners fetch useful data
validators score output
builders get it through an API
That is how a subnet becomes a product.
intotao.app/subnets/desearch
#Bittensor #SN22
2
12
346
Subnet 22 validator update is merged.
Rewards now split 80% AI search, 10% X, 10% Web.
Each type has a quality bar, and real volume matters. Serving only one type no longer earns full credit.
PR: github.com/Desearch-ai/subne…
#Bittensor #SN22 #Desearch
1
9
376
Benchmark v1 is only the start.
The real goal is a stronger measurement layer for decentralized search.
More runs. Better question sets. More categories. More pressure on miners to improve.
Search gets better when competition is measured.
#Bittensor #SN22
1
13
265
We made the benchmark data public.
Questions, answers, sources, scores, and verdicts are available for builders and researchers to inspect.
Dataset:
huggingface.co/datasets/dese…
Better evals create better incentives.
#Bittensor
1
9
357
This benchmark does not use a static answer key.
The judge reads the returned answer, cited pages, and source content.
We score:
Source relevance
Answer quality
Groundedness
Search quality should be measured on what users actually receive.
22.desearch.ai
1
7
240
Source relevance matters because bad sources make good looking answers useless.
In our first benchmark, Desearch posted the strongest source relevance score: 92.3%.
That is the signal we care about: search that finds pages worth citing.
22.desearch.ai
#AISearch #SN22
1
1
12
266
The key idea behind decentralized search:
Miners compete to return better answers and better sources.
Validators measure quality.
Rewards move toward better behavior.
The product improves in public.
That is the loop we are building on Desearch.
#Bittensor #SN22
2
2
14
379
Desearch.ai | Subnet 22 retweeted
Bittensor capturing all the excitement at the Louve this week for @proofoftalk.
Amazing to connect with @GordonFrayne, @SiamKidd, @Raleigh_CA and so many others from the community.
Keep the conversations coming for Day 2.
1
8
56
2,386
Open benchmarks beat black box claims.
We published the Desearch search eval repo so builders, miners, validators, and researchers can inspect how the benchmark works.
Repo:
github.com/Desearch-ai/desea…
#Bittensor #DecentralizedSearch
2
2
11
435
Decentralized search is no longer theoretical.
Desearch led the composite leaderboard against centralized providers in our first benchmark, with the strongest source relevance.
This is v1.
More miners. More competition. Better search.
22.desearch.ai
#Bittensor #SN22
2
4
22
1,471
Benchmark repo:
github.com/Desearch-ai/desea…
Public evals let miners, validators, builders, and researchers inspect the methodology and help improve the benchmark.
1
2
186