As promised - introducing Notice CLM Notice v0.1.0 is an open-source, AI-ready contract lifecycle management platform Governed agent workflows, customer-hosted OCR, DocuSeal integration, auditable evidence, and enterprise controls built for teams that need automation without losing oversight github.com/ericwithac/notice…

@_sigmaX_ // Borrowed Light Some places are known without ever being seen. Emerging from photographs, stories and cultural memory, they take shape on the canvas, where imagined places begin to feel remembered. 📍 StudioG, Rome 🗓️ 25 Sep // 25 Oct MP TransientLabs

Thank you bro! It’s from my NEW YORKERS collection, I made you into an honorary The piece includes a work of mine you’ve previously collected and our interview from my podcast synthesized into a piece of art Would love to know what you think of the rest of the collection N3wyorkers.com 🧿

head chef astra has been cooking

TensorFold vs vLLM, Qwen3.8‑27B, one Spark, same 3 prompts, thinking off: 3× faster than vLLM on the same DGX Spark. Byte-identical output. • TensorFold (DFlash2 drafts): 102.9 tok/s • vLLM MTP=3 (NVFP4): 34.9 tok/s • TensorFold serial, drafts off: 12.9 tok/s Per prompt (TensorFold / vLLM): sequence 126.9 / 37.0 code 75.7 / 32.7 json 106.2 / 35.0 TTFT 0.12 s. Drafted vs serial output: sha256 identical on all three. Loads in 32 s. Not weight-matched: MLX 4‑bit g64 vs NVFP4 ModelOpt. Single pass. Long structured outputs draft well; prose will land lower. Bonus: one Spark now beats my M4 Max on this model (oMLX MTP3: 65.8). github.com/ashhart/TensorFol… — @ashxhart

Can also send him funds via @XMoney

Even tho my nfts weren’t lucky enough to be included I have to give @0xQuit major credit for what he did the other day and his continued support for our community Just donated 🫡

the chance AI will kill us all is close to 0% the chance that Anthropic or OpenAI want to have ultimate control over humanity is close to 100% few understand this.

Update Regarding Unauthorized BCTC Token On September 9 we warned that the BCTC fungible token circulating on Robinhood Chain was not issued, authorized, or endorsed by Blockticity. We have continued investigating and can now provide additional information. Blockticity has never issued a fungible token, created or authorized a liquidity pool or trading pair, or offered a fungible token for sale. On-chain records show that the unauthorized BCTC token was created through a transaction signed by a legacy Blockticity private key that had been compromised in April 2026. Following the compromise, Blockticity retired the key from operational use and migrated its contracts to new keys on April 21, 2026, more than four months before the unauthorized BCTC token was created. Those migrations are publicly verifiable on-chain. The records further show that, shortly before the token was created, a wallet not owned or controlled by Blockticity funded the compromised key with ETH used for transaction fees. The compromised key then signed a transaction calling the launchpad's token-creation function. The same wallet that funded the compromised key subsequently collected fees associated with the token launch. Blockticity did not authorize, fund, promote, or participate in the launch of the BCTC token. Blockticity did not receive or control any proceeds or fees associated with its launch and does not hold any BCTC tokens. For clarity, Blockticity's legitimate on-chain activity includes the issuance of Certificates of Authenticity: ERC-721 records tied to authenticated documents and products. These certificates are not fungible tokens, are not issued or marketed as investments or tradable assets, and Blockticity does not operate a market for them. Our full token policy is available at blockticity.io/token-policy. Anyone who purchased the unauthorized BCTC token may file a report with the FBI's Internet Crime Complaint Center at ic3.gov. We recommend including relevant transaction hashes and wallet addresses when submitting a report. Blockticity has reported this incident to the FBI's Internet Crime Complaint Center, the relevant launchpad, and listing venues. We are continuing to investigate.

Around 4B paper documents circulate through the $25T global trade market, making document fraud a massive, systemic risk. @blockticity is changing that by bringing over 1M authenticated records ($7.7B+ in real-world assets) onto Walrus. By mirroring 15TB+ of trade evidence across five commercial verticals, Blockticity gives partners public, cryptographic proof of origin while keeping sensitive commercial data protected. Read the full announcement: blog.walrus.xyz/blockticity-…

GM. This is what verifiable trade infrastructure looks like. Anyone can verify a document hasn’t been altered, while only authorized parties can read what’s actually in it. No single company holds the key. Not Mysten. Not us. That’s Seal, and it’s legitimately hard to build. Now more than $7.7B in authenticated trade records are live on @WalrusProtocol. Genuinely fun to build on, and the @Mysten_Labs crew have been great to work with! 🩵

Ran out of my weekly credits on my $200/mo OpenAI plan Purchased $200 in additional credits figuring it was roughly the same amount of usage Then blew through the additional credits in less than 30 mins Thought it was just a UI bug but it showed only 2% weekly usage remaining after I purchased the credits Pretty wild… Does this roughly mean the subscription plans are subsidized 50x versus the usage rack rates?

inspired by this, ive been building a fully functional human and agent usable CLM (contract lifecycle management) app bc Ironclad sucks and is super expensive for the value it offers ill be releasing it fully open source this week 👀

One conviction drives everything we do: as AI agents begin running real economic activity, they'll need a foundation they can verify. We have total conviction that web3 rails, products, and solutions will serve as the necessary bedrock for an AI Agent-led economy. We’ve been building through the bear market while thinking deeply about what comes next. Our CEO has distilled this down into a blog post, which we are proud to share today. onesource.io/blog/onesource-…

Recently completed paintings: 🍓🍓 88¢ Berries (1-4), 2026, 12x12", Oil on Canvas 1/

MCDMA | Metal CUDA Direct Memory Access 🚀 If you have a Spark and an Apple Silicon Mac, MCDMA gives you a direct RDMA path between CUDA memory and Metal-side unified memory over USB-C. Registered memory, rkeys, one-sided READ/WRITE, two-sided SEND/RECV with credit flow control. Same verbs both ways, no master/slave. The Mac writes straight into CUDA-mapped memory on the Spark, and the Spark writes straight back into Mac memory. My setup takes it a little further: Spark 1 ⇄ CX7 ⇄ Spark 2 (prompt processing) Spark 1 ⇄ USB-C ⇄ Mac Studio (Decode) Spark 2 ⇄ USB-C ⇄ Mac Studio (Decode) Two independent MCDMA USBC links, so the Studio isn't stuck behind one cable; both Sparks move data concurrently, and it writes results back into either. Measured, every byte delivery verified: • 939 MB/s single link • 1.80 GB/s Mac → both Sparks, concurrent • 1.25 GB/s both Sparks → Mac, concurrent • 24 µs round-trip, 41k msg/s small-message One Spark + One Mac works. Two is just how I'm using it: DeepSeek prompt processing across the Sparks, decode on the Studio. Benchmarks, tests, Open Source, and write-up this week. @NVIDIARTXSpark @NVIDIAAI @NaderLikeLadder @msharmavikram There’s still a lot of performance headroom here. If the currently locked USB4 controller can be allowed to train at full capability, I’d love to test how far we can push this. Please check your DMs.