https://xitter.cf/t.co/372esc0cFD / https://xitter.cf/t.co/zHeHZHJUvS nostr: [email protected]
Joined July 2021
- Tweets618
- Following1.3K
- Followers1.3K
- Likes11.6K
Richard Safier retweeted
kind of like this remotion ai generated video, also a sneak peak what our team is building, let me know if you math-adjacent and want to have a play (i know nothing about math)
Richard Safier retweeted
BREAKING: bitcoin sent via apple's "find my" mesh network
possibly the most absurd medium to send bitcoin yet
a small device convinces nearby iphones to transmit tiny amounts of data with cashu ecash to the recipient.
without any internet connection.
magic i̶n̶t̶e̶r̶n̶e̶t̶ money.
Richard Safier retweeted
New contender?
New 33B param model ranking as high as DS4-Pro, multimodal support, 262k context
Perfect for:
- 3090 / 4090 / 5090 / RTX pro 6000
- supports: high / medium / low reasoning
huggingface.co/Agnes-AI/Agne…
Richard Safier retweeted
Atlas passed 1,000 issues and PRs combined. Thanks to everyone contributing, including @tbraun96 @MarcusAerulius @NodeHello @AlexiDerkatsch.
Follow @atlasinferencex and come build with this great community:
github.com/Avarok-Cybersecur…
Richard Safier retweeted
Turns out that with enough math a counterparty can do multisig in a Lightning channel without changing the protocol: eprint.iacr.org/2026/1757
Richard Safier retweeted
@AtlasInference just took the single-Spark decode crown for Qwen3.8-27B.
30+ tok/s server-measured on one DGX Spark:
32.0 / 29.8 / 30.5 across 3 runs. Deterministic replication across labs, same day.
Same protocol, same checkpoint, same box on vLLM: 28.2. Atlas wins decode with thinking on too (24.1 vs 23.1).
And 30 looks like the floor, not the ceiling: this engine's DFlash drafts already hit 37.5 tok/s on Qwen3.6, the 3.8 drafter is in training, and a prefill fix showing +80–165% is in test. The Spark decode ceiling is moving weekly.
github.com/Weschera/Atlas-DG…
Richard Safier retweeted
Excited to present Atlas TUI for Edge Hardware (DGX Spark/Strix Halo)
Beautifully polished interface with built-in evaluation harnesses, hot model swaps, and comprehensive metrics ✅
We made sure Atlas was easy to spin-up before, now you can verify and chat models in-place 🔥
Richard Safier retweeted
Coldcard Dicerolls: Verification Report
I spent a lot of time today verifying the Coldcard Diceroll seed path in multiple firmware versions. The implementation is correct and can generate up to 256 bits of entropy with 100 dice rolls.
Full report below.
github.com/portlandhodl/cold…
Richard Safier retweeted
MK4s are vulenerable but are millions of times harder to steal from. MK3s can be stolen from a decent laptop, mk4,5s and Q would need much more compute.
The model:
- attacker has a roughly 2^20 UUID range that it knows cold cards fall in
- checks a 16 button press variancer in user behavior before generating the seed
You should still promptly remove funds from mk4,mk5 and Q.
Richard Safier retweeted
Pleased to share Atlas Inference is now part of @NVIDIAAI's Inception program 💚
Excited for the technical resources/networking opportunities, gtm support, and sweet pricing on hardware 💻
AI workloads keep scaling, we strive to have optimized inference for all enthusiasts 🌎
Richard Safier retweeted
👨🏻🏫 New Bitcoin Whiteboard: State of bridging and ZK verification on bitcoin. 👨🏻🏫
Chapters
00:00 Overview
06:02 Garbled Circuits Magic
11:17 Supporting Multi-Chain.
12:34 Added Trust Cross-chain (Bridging To Ethereum). 15:00 Security Council
Richard Safier retweeted
We're excited to share we're upgrading the dual sparks to the "trifecta" setup🔼 build.nvidia.com/spark/conne…
What models run best on 3 GB10s (384GB unified)? Are there any models to be better optimized with Atlas for this ring topology?
Maybe Nemotron-Ultra, GLM-5.2, LET US KNOW!
Richard Safier retweeted
Humans are impatient. We get that ⏰
If you gave up on Atlas because of waiting, just try again. It'll take <90s 🏎️
NO ONE should wait on inference. We solved our TTFT woes. If ANYONE serves Qwen3.6-27B-NVFP4 better than Atlas Inference, TELL US!!! 📭
Run command attached 🧵
Richard Safier retweeted
Today I'm happy to announce Noah. ⛵
It's a Bitcoin wallet that @nitesh_btc and I have been working on, using the Bitcoin Ark protocol via @secondhq.
Noah and the Ark protocol allow easy onboarding, Lightning Address support, easy backups, fast payments, and much more, while staying trust-minimized.
With Noah, you can easily pay and receive via Lightning without dealing with channel management and inbound/outbound liquidity constraints.
Onboarding is as easy as receiving a Lightning payment. When you onboard, you get an Ark VTXO (akin to a UTXO but offchain) instead of a channel.
You can also onboard via onchain and internal Ark payments.
Each Noah user gets a Lightning Address that they can use to receive, even offline for Noah->Noah payments.
You can restore the wallet securely with just your seed phrase.
Noah is available now on mainnet! Please give it a shot and let us know what you think. ⚡️
