67 @67apay

Hashmind is live on @Ethereum. $HASH 0xe7802782ac7679b4ff0e7106b3230adfef251688 Post-quantum AI agents. Verifiable by design. Every agent signs its decisions with hash-based keys. Uniswap v4 hooks verify those signatures onchain. No blind trust. Every move authenticated. hashmind.fun

Nock Agent update. We ran Gemma 4 31B, our base model, through our first agent test. It solved 30 of 30. Sounds great, but it means the test was too easy. Training can’t beat a perfect score, so the test couldn’t show us anything. So we built a harder one: 20 tasks, with hidden tests the model never sees and traps for the lazy fix. Base score on it comes next.

We're partnering with @useNockAI. NockAI makes Nock Coder, an open coding model small enough to run on a laptop. It is coming to Axon. Here is what that means. When you build on Axon, you will be able to pick Nock Coder as the model for small code jobs: writing a function, fixing a bug, turning an idea into a snippet. It runs through Surplus at $0.03 per million tokens, so code jobs on Axon get far cheaper. For NockAI, it puts their model to work on real jobs from day one, in front of every builder on Axon. Integration coming soon.

Glad to be working with @axon402! Small code jobs don’t need a giant model. Nock Coder is built for exactly that, and now it gets real jobs from Axon builders. We’ll publish how it does, misses included.

Axon @axon402

2 days ago

We're partnering with @useNockAI. NockAI makes Nock Coder, an open coding model small enough to run on a laptop. It is coming to Axon. Here is what that means. When you build on Axon, you will be able to pick Nock Coder as the model for small code jobs: writing a function,

2 5 15 1K 0

First $NOCKAI burn: 4.56M tokens sent to the dead address. This one was done by hand. We’re preparing a flywheel to make it automatic: 70% of the creator fees we receive for compute and the next model, 30% to buy back and burn $NOCKAI.

Before training Nock Agent, we wrote the exam. 30 small Python projects: 15 bugs, 8 features, 4 multi-file fixes, 3 exact error messages. A task counts only if the tests pass and the test file is unchanged. Two models already tried to pass by editing the test. Training data never overlaps it. Base model first, then the trained one.

Nock Coder answers. The next one acts. Nock Agent is built on @Google Gemma 4 31B. We’re training it to read the code, make the edit, run the tests, and stop when they pass. It goes on Surplus when it’s ready. Every test result goes public, before and after training.

Worth adding the other side of the market: the models. Open fine-tunes are mostly made for free, on personal GPUs. @SurplusRH turns that around: launch the model, and its fees fund the GPU that serves it. We’re one of the four models already running that way. The next one is 31B.

0xJeff @0xJeff

3 days ago

Inference teams worth keeping an eye on this cycle actual computer (SN95) antseed blockrun chutes (SN64) darkbloom dolphin ai engy (SN53) gm (SN28) morpheus orbio surplus intelligence usepod venice ai warden/holo Anyone else?

52 19 225 23K 178

The four, from 0 to 4: Nock Coder 1.5B and its Qwen2.5-Coder base wrote every tool call as plain text, so nothing ran. Qwen3 1.7B solved 1, and once edited a test file instead of the code. Qwen3 4B solved 4, about 5 seconds each. Only 5 tasks, so it’s a first look. A 30 task run is next.

We tested this. Same harness, same tools, same 5 tasks, four open models. Tasks solved: 0, 0, 1 and 4. The harness matters. At small sizes, the model still decides.

Mr. Buzzoni @Mr_Buzzoni

a week ago

I'm sharing with you one of the best papers I've read on Harness Engineering it takes apart Claude Code (Opus 5.5), Codex, Gemini CLI and more, and boils all of them down to one line: agent = model + harness what you walk away with: > what a harness actually is > how today's

67 250 1K 83K 2K

Glad to be one of the four. Nock Coder is live on the Surplus API, and the live demo on nockai.dev runs on it. Room for bigger models is good news for what we’re testing next: an agent model that edits files, not just answers. Soon.

Surplus @SurplusRH

3 days ago

Thanks @MCGlive and @ChillTRD for breaking down the thesis. The short version: inference gets a market price, and open models get a launchpad that pays for their own GPUs. What's live today: → 4 open models launched @Vela_LLM @LittleBit_llm @Artex_llm @useNockAI , all 4

7 6 47 6K 3

Nock Coder now works inside @code. Install the @ollama extension, open Chat, pick nockai/nock-coder. Ask about your code and the answer starts with the code. 986 MB on your laptop. No API key. Works with the Wi-Fi off. Guide: nockai.dev/guides/vscode

What actually happens when you ask Artex a question? One real request, step by step: browser → Cloudflare Worker → Surplus API → artex-coder-7b, pinned to the same commit as on Hugging Face. 1.65s to first byte. 4.77s total. One regex, one sentence.

Quick reminder, three ways to use Artex today: 1. Try it: artexllm.com, free, no sign-up 2. Run it: 4 GB MLX build, 63 tok/s on an M1 Max 3. Call it: Surplus API, $0.03 per 1M tokens Same open weights all three ways.

Same question. Two models. The base model writes 140 words. Nock Coder writes 23, and the first line is code. Across all 50 test prompts: 50 words on average vs 185. Both answers unedited. All of them are public: nockai.dev/evaluation

5.9 GB for agent behavior is a great deal. The other end of the range: 986 MB for quick, code first answers when you don't need an agent. Runs fine on a CPU. > ollama run nockai/nock-coder

David Hendrickson @TeksEdge

5 days ago

Kimi K3 is a 2.8-trillion-parameter model. Someone is distilled its coding-agent behavior into Qwen3.5-9B and the Q4 GGUF is 5.9GB. Not Kimi's 2.8T parameters but its behavior. The training set captures 582 real Kimi K3 agent trajectories: 🔎 inspect repos 🛠️ call tools 🧪 run

9 8 147 10K 134

Nock Coder is now in the @ollama library. > ollama run nockai/nock-coder One command, 986 MB, and it answers with the code first. No account, no API key, no GPU needed.

Running it locally gets you v2.1 now. Already have v2? Run the pull command again. The live demo on nockai.dev and the @SurplusRH API still serve v2, the commit we launched with. > ollama pull hf.co/Nock-AI/nock-c…

One number went down. Solidity: 10 of 15 answers compile, v2 had 11. Every score and every raw answer is on nockai.dev/evaluation.

Same data as v2. The difference is how hard we trained. v2: 2 epochs at a higher learning rate. It learned the short, code first style but forgot some Python along the way. v2.1: 1 epoch at half the rate. It keeps the style and the skill.