ocdbytes

Four hikes in Patagonia this week: Laguna de los Tres, Laguna Torre, Laguna Esmeralda, and Mirador de las Águilas. Two were brutal. El Chaltén had almost no internet, but honestly that was a blessing. No social feeds, no CT—just cards, beers, and hanging out with the people you met on the trail. Next stop: Devconnect 👨‍💻

Introducing Zinc+, where we tackle the problem of arithmetizing and proving computations unfriendly to finite fields. Examples: classic hashes, hash + signature, lattice ops., etc. We prove 7 SHA-256 compressions followed by the ECDSA MSM with:

Give a man an LLM and he'll know everything. Give a man 2 LLMs and he won't be sure of anything.

The EU @EUdigitalID picked @NoirLang for their ZK identity app, and you should too! github.com/eu-digital-ide…

We’ve been talking about it. Now you can explore it. Private. Instant. Compliant. Check it out! perimeter.finance

If you're an AI power user, get prepared. As @0xSero says, the math for subs doesn't add up. $200/mo can't cover the 16+ GPUs needed for power users. So what happens next? → Prices go up → Limits get tighter → Your data becomes the product But there’s another path emerging. Open-source models are coming out with incredible efficiency savings. Local inference is improving. Home hardware is becoming viable. The frontier is shifting from access → ownership. Start building your sovereignty now.

We've raised $2.1M to fix your focus. Our wearable headset @mavehealth improves attention & stress regulation in just 20 minutes a day for users at @Google, @ufc, @ycombinator. Backed by @BlumeVentures, alongside existing and new investors. Order now at mavehealth.com 🇺🇸🇮🇳

@kirat_tw I manually write the code for components I find interesting or are new for me, otherwise delegating to claude-code is better as it can deal with boring parts. win win

Curious how has coding changed for you in the last ~1 year. Is anyone using VSCode anymore? Cursor? Claude code? How much of your day is spent coding vs prompting. What do you do when the AI is coding? Do you parallelly start multiple tasks? Are you more productive?

Compliance isn’t optional. Perimeter builds it in from day one!

Compliant by design. Built for the real world.

I ❤️ @claudeai

SHA256 is everywhere in crypto, but inside zk circuits it becomes one of the biggest bottlenecks. Most implementations pay a huge cost because SHA256 was never designed to be circuit-friendly. In our latest write-up by @0xrosetteeee, we explore how to make SHA256 significantly cheaper in R1CS. Key ideas: • Spread-based encoding for bitwise ops • Dynamic bit-width optimization • Single-constraint multi-operand additions • LogUp batching and micro-optimizations This design achieves state-of-the-art SHA256 compression in R1CS among existing open-source implementations. This is particularly important for mobile proving environments, where witness size directly impacts memory usage. Full deep dive ↓

We added WebGPU to Jolt's WASM prover. Result: 2x faster proving. Same proof. Proving runs on your GPU through WebGPU. Mac + Chromium supported right now. These are the initial numbers we're seeing. 2x is where we're starting. The numbers vary depending on the chip. M4: 2x faster M4 Pro: 2.75x faster There's a long list of optimizations we haven't touched yet (SIMD being one) that could push this to 3x and beyond without much effort. Under the hood: Jolt uses an elliptic curve based polynomial commitment scheme, Dory. MSM and pairing operations are the best candidates for GPU acceleration. We plugged WebGPU into Jolt's WASM implementation to offload those ops to the GPU. The Jolt WASM prover was announced at ETHDenver just two weeks ago. We already have GPU acceleration running on it through WebGPU: youtube.com/watch?v=a6y4ZA… This was about getting WebGPU integrated end to end into Jolt's WASM prover and shipping it. These are the first real numbers. Although, there's a lot of headroom left. Will keep posting numbers as we optimize further. Go try it, break it, tell us what you see → joltprove.atheon.xyz

BOOM! Apple’s Neural Engine Was Just Cracked Open, The Future of AI Training Just Change And Zero-Human Company Is Already Testing It! In a jaw-dropping open-source breakthrough, a lone developer has done what Apple said was impossible: full neural network training– including backpropagation – directly on the Apple Neural Engine (ANE). No CoreML, no Metal, no GPU. Pure, blazing ANE silicon. The project (github.com/maderix/ANE) delivers a single transformer layer (dim=768, seq=512) in just 9.3 ms per step at 1.78 TFLOPS sustained with only 11.2% ANE utilization on an M4 chip. That’s the same idle chip sitting in millions of Mac minis, MacBooks, and iMacs right now. Translation? Your desktop just became a hyper-efficient AI supercomputer. The numbers are insane: M4 ANE hits roughly 6.6 TFLOPS per watt – 80 times more efficient than an NVIDIA A100. Real-world throughput crushes Apple’s own “38 TOPS” marketing claims. And because it sips power like a phone, you can train 24/7 without melting your electricity bill or the planet. At The Zero-Human Company, we’re not waiting. We are testing this right now on real ZHC workloads. This is the missing piece we’ve been chasing for our Zero Human Company vision: reviving archived data into fully autonomous AI systems with zero human overhead. This is world-changing. For the first time, anyone with a Mac can fine-tune, train, or iterate massive models locally, privately, and at a fraction of the cost of cloud GPUs. No more renting $40,000 A100 clusters. No more waiting in queues. No more massive carbon footprints. Training costs that used to run into the tens or hundreds of thousands of dollars? Plummeting toward pennies on the dollar – mostly just the electricity your Mac was already using while it sat idle. The AI revolution just moved from billion-dollar data centers to your desk. WE WILL HAVE A NEW ZERO-HUMAN COMPANY @ HOME wage for equipped Macs that will be up to 100x more income for the owner! We’re only at the beginning (single-layer today, full models tomorrow), but the door is wide open. Ultra-cheap, on-device training is here. The future isn’t coming. It’s already running on your Mac. Welcome to the Zero-Human Company era.

Introducing eigen-skills and eigen-mcp🏎️🛠️ , 2 npm packages that give AI agents live access to the full @eigencloud stack. > Six skills covering restaking, AVS, rewards, delegation, EigenCompute TEE, and EigenDA blob storage. Each skill is a SKILL.md file, the agent reads it, matches user intent, and runs the right API call > npm install eigen-skills and Claude Code discovers it automatically. A postinstall script copies the SKILL.md into .claude/skills/. > Set your API key and the agent can query operators, stakers, TVL, APY — all real-time from EigenExplorer > eigen-mcp wraps the same 6 JS clients into 21 MCP tools with Zod-validated schemas. One server, STDIO transport, works with Claude Desktop, Cursor, Windsurf or any MCP Client (Link in comments⬇️)

Private by default. Coming soon.

@Chiragagrwal7 Can vouch for the og personally

in your 20s or 30s you will find me it’s very important you hire me and take your product to the next level

@zeeshan_utd @eigencloud W.

Introducing News Analyst — a fully verifiable AI engine powered by @eigencloud suite > Uses EigenAI with wallet-signed grants. Temperature locked to zero with fixed entropy seeds. No hallucinations, only deterministic truth. > Every analysis is dispersed as a blob to EigenDA. No database—the state is stored across a validator set on Sepolia. > Generate a DA Commitment for every report. Reconstruct the entire analysis purely from the sharded network. > Sidecar architecture running via Dockerized EigenDA Proxy.

5.3-Codex did for research engineering what 4.5 Opus did for front-end engineering. Don't try it on the problem you're working on—try it on the problem you thought was too big to solve

Compromised and revoked TEE machines could pass dstack's attestation verification as perfectly valid, due to missing checks. What's more? This gap has existed since the library's first commit. @PhalaNetwork Cloud and every protocol built on it inherited this behaviour from day one. Their GHSA marks this as Critical and notes that it "bypasses entire remote attestation model". My team at @bluethroat_labs reported this and 5 other vulnerabilities, and this is the response we got: + $2,500 in bounty offered + disclosure timelines framed as "threat" + wiped shared Notion + severities downgraded in a public blog post Here's the full story: 🧵👇🏻

Banks take days to settle. Crypto is instant but public. There's a third way.