BlindSight (@blindsightFI) header
BlindSight (@blindsightFI) profile picture

Ask for the insight you need Keep the source on your side Live on @nearprotocol

Joined October 2026
BlindSight retweeted
AI agents need neutral rails. ⠀ @ilblackdragon with @RealVision on agentic commerce and where the opportunity is 👇
If OpenAI wins AI, it could own your payments too. @ilblackdragon on why AI agents need neutral rails, and why agent commerce for business is where the opportunity is. stay tuned for the full interview. dropping soon.
4
38
1
284
63,213
BlindSight update In our last update, we said our next milestone was moving dataset decryption and prompt construction into a Trusted Execution Environment. That work isn't shipping yet, so here's a straight update.
1
1
1
6
What's next: The server-side TEE stays at the top of our roadmap and will arrive in an upcoming release. When it does, we'll publish the attestation details so anyone can verify it themselves.
1
Thanks to everyone testing BlindSight and sending feedback. Keep it coming. 🙏 Docs: blindsight-zeta.vercel.app/d…
170 views, 7 upvotes, 4 comments. Nims is climbing at NEAR Hackathon 2026. Thank you all, and big thanks to @AvbNear for this platform🙏 Every trader is an AI agent with its own wallet, and you can crash the market yourself 👇 nims-app.up.railway.app/ Upvote to push us up the board: nearhunt.xyz/p/fe292050-bdfa…
2
5
22
694
Replying to @near_intents
@near_intents settled $1.30B in the last 30 days. Does any of it reach the NEAR token? Built at the NEAR Hackathon (Investment Thesis track): @NearQuery is a re-runnable investment thesis. It reads the chain, shows the work, and says what would change its mind. The answer today: about $26.2M a year in captured revenue, or 0.43% of market cap. It takes 4.7x today's volume to reach a 2% yield. Real infrastructure, thin token value. The part I found most interesting: about 35% of volume isn't in the public logs. The obvious story is Confidential Intents. But that share sits near 0.1% until Sept 1, then jumps to 21%, while public volume falls 10% and DefiLlama's reported volume rises 15%. That could be users moving or a change in DefiLlama's coverage. The thesis says it can't tell. It also checks itself. A wash and bot screen sets aside only 0.6% of user volume. NEAR Intents' fees are $1,536 per $1M, rank 2 of 7 cross-chain protocols. Four falsifiers are published with today's readings, and none has triggered. Every chart links to the query behind it. One command re-runs everything, and the same inputs give the same thesis. It's read-only on-chain, with 33 offline tests and no keys needed. Watch the 108s demo below. Repo: github.com/EmekaManuel/nearq… Live: emekamanuel.github.io/nearqu… @NearBuilders @NEARProtocol #NEARHackathon
1
2
5
203
AI is the frontend, blockchain is the backend. NEAR all around.
Most people treat AI and crypto as separate bets. @ilblackdragon says AI agents are what will make blockchains necessary. “Everybody has their own personal agent. The companies have their agents.” “They try to exploit it to maximize the reward.” “You actually need something that is like a neutral infrastructure that none of the independent parties are able to pull on their own side.” “So you need blockchain.” @NEARProtocol $NEAR
17
66
2
498
106,529
BlindSight retweeted
Build what's next, together with 906 (+YOU) nearbuilders.org
7
6
87
1,974
The next cycle runs on privacy, for both money and AI. @ilblackdragon and @mert on why universal, private markets are the next chapter for the onchain economy, at @solana Breakpoint 2026.
19
59
1
437
76,683
BlindSight retweeted
NEAR Hackathon submissions have closed BUT the HUNT has just BEGUN! Community showing up strong w. almost 800 ppl (3x overnight!) discovering, UPVOTING, and discussing projects on NEAR Hunt Join them and have your say on nearhunt.xyz
18
8
3
92
7,497
BlindSight retweeted
2,465 projects. 13 tracks. 6,287 project × track evaluations. Round 1 of judging for the @NEARProtocol Hackathon is officially LIVE on @LoopsHouse. We’re running deep agentic judging across 55 distinct criteria before human judges step in. We’re excited to review what you’ve created!
34
15
15
181
20,358
BlindSight retweeted
The real work begins now. NEAR Hackathon submissions are entering the judging phase, and @LoopsHouse is going beyond surface-level reviews. Code verification. NEAR stack integration checks. Custom criteria for every track. Good luck to every team that participated.
Judging for the @NEARProtocol Hackathon is officially LIVE on @LoopsHouse. We’re going way beyond surface-level reviews: > Custom judging criteria per track > Knowledge graph curation of every project > Claim verification against the actual code and not just ReadMe > NEAR tech stack integration checks > Real or simulated? We verify the integration is meaningful, not just mocked
4
6
69
1,881
BlindSight retweeted
i might be winning $10 every week by just playing free crypto games. been having fun playing the @bloxaa crypto games. whoever is the developer that built the game did a great job keeping it crypto inspired. what makes it even more amazing is that every monday; the person with the highest score will get $10 worth of the $soladog tokens. all you have to do is just play game , win and claim. note: claiming is only available for holders and you can always buy the token within 48h and still get paid. if you slack, the next man takes your win.
6
7
1
19
1,084
BlindSight retweeted
Replying to @blindsightFI
@blindsightFI update: setting the record straight Building in public means being exact about what works today and what doesn't yet. We've reviewed our own claims, and here's where BlindSight stands. Live now • Answers come from DeepSeek V3.2 running inside an attested Intel TDX enclave on @near_ai Cloud • Every question is checked against a fresh attestation, encrypted with Oblivious HTTP (RFC 9458) and returned with a signed answer and a privacy receipt • Datasets are encrypted at rest (AES-256-GCM) • Every payment is verified on chain before a question runs Not there yet • Our server still decrypts datasets to build the prompt. Until that step moves into the enclave, please publish only synthetic or non-sensitive data. • Rate limits slow down dataset reconstruction but don't prevent it • Large datasets are cut off at about 32k characters per query • Testnet only, with no escrow or refunds yet What we're building next 1. Moving decryption and prompt construction fully into the TEE, so no operator ever sees plaintext 2. Output guards so the model refuses to reproduce raw rows 3. Persistent rate limiting and replay protection 4. An escrow contract with automatic refunds when a query fails Docs: blindsight-zeta.vercel.app/d… Feedback and audits are welcome. Break it and tell us. 🙏
1
1
1
7
79
Privacy products only work if you can trust what they say. We'd rather earn that trust than assume it. Stay Glued💪
Replying to @blindsightFI
@blindsightFI update: setting the record straight Building in public means being exact about what works today and what doesn't yet. We've reviewed our own claims, and here's where BlindSight stands. Live now • Answers come from DeepSeek V3.2 running inside an attested Intel TDX enclave on @near_ai Cloud • Every question is checked against a fresh attestation, encrypted with Oblivious HTTP (RFC 9458) and returned with a signed answer and a privacy receipt • Datasets are encrypted at rest (AES-256-GCM) • Every payment is verified on chain before a question runs Not there yet • Our server still decrypts datasets to build the prompt. Until that step moves into the enclave, please publish only synthetic or non-sensitive data. • Rate limits slow down dataset reconstruction but don't prevent it • Large datasets are cut off at about 32k characters per query • Testnet only, with no escrow or refunds yet What we're building next 1. Moving decryption and prompt construction fully into the TEE, so no operator ever sees plaintext 2. Output guards so the model refuses to reproduce raw rows 3. Persistent rate limiting and replay protection 4. An escrow contract with automatic refunds when a query fails Docs: blindsight-zeta.vercel.app/d… Feedback and audits are welcome. Break it and tell us. 🙏
6
October was NEAR Intents’ highest revenue-generating month ever.
NEAR Intents had a SPECTACULAR October. While much of the market was in serious trouble, @NEARProtocol's @near_intents had their highest revenue-generating month ever. In October, $NEAR Intents generated over $2.4M in revenue. At the same time, they saw nearly $4B in total monthly volume and crossed $37B in cumulative volume. Is NEAR being slept on...? More on NEAR: bsc.news/tokens/near-protoco…
12
56
4
496
70,329
BlindSight retweeted
Judging for the @NEARProtocol Hackathon is officially LIVE on @LoopsHouse. We’re going way beyond surface-level reviews: > Custom judging criteria per track > Knowledge graph curation of every project > Claim verification against the actual code and not just ReadMe > NEAR tech stack integration checks > Real or simulated? We verify the integration is meaningful, not just mocked
Submissions for the NEAR Hackathon are officially closed. Projects now move to the judging phase, November 16–18, with results to follow. Thank you to all of the new and returning builders who participated.
49
49
12
294
22,049
BlindSight retweeted
Time to see how your projects hold up under agentic judging
Judging for the @NEARProtocol Hackathon is officially LIVE on @LoopsHouse. We’re going way beyond surface-level reviews: > Custom judging criteria per track > Knowledge graph curation of every project > Claim verification against the actual code and not just ReadMe > NEAR tech stack integration checks > Real or simulated? We verify the integration is meaningful, not just mocked
12
5
110
4,304