Pu.sh
I originally was just messing with pi-autoresearch. Gave it a sample task to build the most portable coding agent. First cut was 6 KB of shell. Great for one-shots, unusable interactively. I was shocked it actually worked. Started building up -- adding features — but with a self-imposed rule: no new dependencies, and sub 500 LOC. This thing had to be truly portable. Just sh, curl, awk. System primitives only. Which means I did some genuinely disgusting things in awk, including JSON parsing and the OpenAI Responses tool loop with reasoning items carried across turns. It's now ~400 lines. In the box: Anthropic + OpenAI, 7 tools (bash, read, write, edit, grep, find, ls), REPL, auto-compaction, checkpoint/resume, pipe mode, 90 no-API tests. Not in the box: TUI, streaming, images, OAuth, Windows, dignity. Two honest things: 1. I stole/modified the system prompt and the architecture. Pi/Claude/Codex wrote the awk. I cannot read most of this code. This wasn't possible for me a year ago. 2. Heavily inspired by Pi (pi.dev) — same 7-tool surface, same exact-text edit model. Credit where it's due. Pi is awesome -- you should probably use them. The agent loop itself is tiny. Almost everything else in a "real" agent CLI is DX and hardening. You can probably build your own harness exactly how you like it. Mario Zechner's AI Engineer talk on taking back control of your tools nudged me here. The name is because it's a .sh file. The other thing it sounds like is, regrettably, also accurate.
Spec27
Hi HN! We’re a team of ML validation specialists and we’ve been building /Spec27, a tool for testing whether AI agents still do their job safely and reliably as models, prompts, tools, and surrounding systems change. We started working on this because a lot of current LLM evaluation work seems aimed at scoring general model behavior, while many teams are deploying systems that have a specific mission to fulfill. Many of the tools also assume you have full access to the agent stack and traces so you can place SDKs and Gateways, but a lot of agents are being created on vendor platforms where this isn’t possible. As a result, we approaches it from the outside in: all tests just run to the primary interfaces of an Agent and don’t assume anything about internals. The other important things about the approach is spec-driven. Instead of treating testing as a one-off benchmark or static eval set, we let teams define reusable specifications for the behavior they want from an agent, then generate tests against those specs. With this you can automatically generate adversarial and robustness checks, so you can see what an agent is sensitive to and what kinds of changes cause it to fail. We’ve worked on validation for other AI systems before, including vision and tabular workflows, and /Spec27 is our new product for language-model-based agents. Currently in early access, so we’d love feedback! The current version is strongest for single-turn agent and application validation. We do not fully support multi-turn interactions yet, and better telemetry/tool-call integration is still on our roadmap. We’ve made the product open to try for HN readers, with a sample flow so it’s easy to poke around without much setup. We’d especially love feedback from people deploying internal agents, vendor agents, or other AI systems where reliability matters more than benchmark scores.
The Dominion List
Show HN: The Dominion List – an open-source db of Canadian founders in the US
Filling PDF forms with AI using client-side tool calling
Hey HN! I built SimplePDF Copilot: an AI assistant that can interact with the PDF editor. It fills fields, answers questions, focuses on a specific field, adds fields, deletes pages, and so on. It's built on top of SimplePDF that I started 7 years ago, pioneering privacy-respecting client-side pdf editing, now used monthly by 200k+ people. As for the privacy model: the PDF itself never leaves the browser. Parsing, rendering, and field detection all run client-side. The text the model needs (and your messages) goes to whatever LLM you point at. By default that's our demo proxy (DeepSeek V4 Flash, rate-capped), but you can BYOK and point it at any cloud provider, or go fully local (I've been testing with LM Studio). Unlike the existing "Chat with PDF" tools that only retrieve the text/OCR layer, Copilot can act on the PDF: filling fields, adding fields (detected client-side using CommonForms by Joe Barrow [1], jbarrow on HN with some post-processing heuristics I added on top), focusing on fields, deleting pages, and so on. I built this because SimplePDF is mostly used by healthcare customers where document privacy is paramount, and I wanted an AI experience that didn't require shipping PII to a third party. Stack is pretty standard: - Tanstack Start - AI SDK from Vercel - Tailwind (I personally prefer CSS modules, I'm old-school but the goal since I open source it, I figured that Tailwind would be a better fit) The more interesting part is the client-side tool calling: events are passed back and forth via iframe postMessage. If you're not familiar with "tool calling" and "client-side tool calling", a quick primer: Tool calling is what LLMs use to take actions. When Claude runs grep or ls, or hits an MCP server, those are tool calls. Client-side tool calling means the intent to call a tool comes from the LLM, but the execution happens in the browser. That matters for: speed, you can't go faster than client-to-client operations and also gives you the ability to limit the data you expose to the LLM. For the demo I do feed the content of the document to the LLM, but that connection could be severed as simply as removing the tool that exposes the content data. The demo is fully open source, available on Github [2] and the demo is the same as the link of this post [3] What's not open source is SimplePDF itself (loaded as the iframe). I could talk on and on about this, let me know if you have any questions, anything goes! [1] https://github.com/jbarrow/commonforms [2] https://github.com/SimplePDF/simplepdf-embed/tree/main/copil... [3] https://copilot.simplepdf.com/?share=a7d00ad073c75a75d493228...
zot
Why I Built Another coding agent harness?: https://dev.to/patriceckhart/zot-why-i-built-another-coding-... Github Repo: https://github.com/patriceckhart/zot
Utilyze
The standard GPU utilization metric reported by nvidia-smi, nvtop, Weights & Biases, Amazon CloudWatch, Google Cloud Monitoring, and Azure Monitor is highly misleading. It reports the fraction of time that any kernel is running on the GPU, which means a GPU can report 100% utilization even if only a small portion of its compute capacity is actually being used. In practice, we've seen workloads with ~1–10% real compute throughput while dashboards show 100%. This becomes a problem when teams rely on that metric for capacity planning or optimization decisions, it can make underutilized systems look saturated. We're releasing an open-source (Apache 2.0) tool, Utilyze, to measure GPU utilization differently. It samples hardware performance counters and reports compute and memory throughput relative to the hardware's theoretical limits. It also estimates an attainable utilization ceiling for a given workload. GitHub link: https://github.com/systalyze/utilyze We'd love to hear your thoughts!
Tiao, A two-player turn-based board game
Hi HN, I built this digital version of Tiao, a two-player turn based strategy board game. Think Checkers meets Go. It's free, runs in the browser, has multiplayer, AI, over the board mode and a lot of other neat things. The source is on GitHub (AGPL). The game was originally designed by my friend Andreas Edmeier. He created the rules and has been playtesting and refining the game design for years. I built the website for it. The core in about 2 weeks using TypeScript, Next.js, Express, Websockets, and MongoDB. Fully dockerized, deployed on a Hetzner VPS with Coolify. Authentication with better-auth. Real-time gameplay, ELO matchmaking, OpenPanel analytics, and a fully functional achievements system. Play it: https://playtiao.com Source: https://github.com/trebeljahr/tiao Happy to answer questions about the tech, the game design, or anything else. My hope is that more people will play this game because I think it is genuinely fun and would be cool to one day see people play this on a Go board or on their phones/computers. Have a good one.
Physics-accurate 3D assets for robotics simulations from any input
In the past 4 years, we developed a platform generating 3D visuals and deploying AR Try-on experiences into large footwear and retail companies. We focused on high-fidelity meshes and textures since the point was giving shoppers a realistic visual. Then robotics companies started to approach us and they asked for thousands of 3D assets that they can use in creating different simulation environments. This is crucial for them so they can achieve better sim-to-real transfer with domain randomization (train a humanoid in a million different kitchens so a real one is not a surprise when it is deployed in the real world). However, we realized nice geometry and pretty textures are not enough for sim engines as they lack physics properties. So we developed a new pipeline creating SimReady assets in OpenUSD and MJCF that plug directly into Isaac Sim and MuJoCo. - Mass: volume from the mesh, density from material classification by a vision-language model (approach adapted from NeRF2Physics), multiplied together. - Center of mass, static/dynamic friction coefficients and restitution from the same material classification and estimation priors. - Collision meshes via CoACD; adaptive hull count is in progress. - Geometry and texture optimization, plus file format conversions. - We are currently focused on physics property estimation for props. Next step: articulated objects It comes with free credits on easy Google sign-up so you can give it a test. No credit cards required. Still early. All we'd ask is honest feedback. What works, what doesn't, what you wish it did differently. That's worth more to us than anything right now.
Spectrum
how can we make AI agents more accessible? it has been a crazy journey seeing how AI agents gets smarter every single day but interaction layer is still missing we are launching Spectrum TODAY: one Unified API connect your AI Agents into iMessage, WhatsApp, Telegram and more
Graph Compose
Hey HN. Graph Compose is a hosted platform for orchestrating API workflows on Temporal. You define workflows as graphs of nodes (HTTP calls, AI agents, iterators, error boundaries) and everything runs as a durable Temporal workflow under the hood. Three ways to build the same graph: a React Flow visual builder, a typed TypeScript SDK (@graph-compose/client), and an AI assistant that turns plain English into a graph. Open-core: the execution foundations and integrations service are AGPL-3.0. The platform orchestrator, visual builder, and AI assistant are proprietary. Longer backstory on why I built this in the first comment. Would love feedback, especially from anyone who's dealt with the "services work fine, the glue between them doesn't" problem. Docs: https://graphcompose.io/docs
A lightweight way to make agents talk without paying for API usage
Show HN: A lightweight way to make agents talk without paying for API usage
Solyto
Show HN: Solyto – a free, open-source all-in-one personal management app