🛠️ All DevTools

[Other] Rebasing in Magit

[Other] Show HN: How I topped the HuggingFace open LLM leaderboard on two gaming GPUs I found that duplicating a specific block of 7 middle layers in Qwen2-72B, without modifying any weights, improved performance across all Open LLM Leaderboard benchmarks and took #1. As of 2026, the top 4 models on that leaderboard are still descendants.<p>The weird finding: single-layer duplication does nothing. Too few layers, nothing. Too many, it gets worse. Only circuit-sized blocks of ~7 layers work. This suggests pretraining carves out discrete functional circuits in the layer stack that only work when preserved whole.<p>The whole thing was developed on 2x RTX 4090s in my basement. I&#x27;m now running current models (GLM-4.7, Qwen3.5, MiniMax M2.5) on a dual GH200 rig (see my other post). Code and new models coming soon.<p>Happy to answer questions.

[CLI Tool] Show HN: Smux – Terminal Multiplexer built for AI agents

[Other] Show HN: DD Photos – open-source photo album site generator (Go and SvelteKit) I was frustrated with photo sharing sites. Apple&#x27;s iCloud shared albums take 20+ seconds to load, and everything else comes with ads, cumbersome UIs, or social media distractions. I just want to share photos with friends and family: fast, mobile-friendly, distraction-free.<p>So I built DD Photos. You export photos from whatever you already use (Lightroom, Apple Photos, etc.) into folders, run `photogen` (a Go CLI) to resize them to WebP and generate JSON indexes, then deploy the SvelteKit static site anywhere that serves files. Apache, S3, whatever. No server-side code, no database.<p>Built over several weeks with heavy use of Claude Code, which I found genuinely useful for this kind of full-stack project spanning Go, SvelteKit&#x2F;TypeScript, Apache config, Docker, and Playwright tests. Happy to discuss that experience too.<p>Live example: <a href="https:&#x2F;&#x2F;photos.donohoe.info" rel="nofollow">https:&#x2F;&#x2F;photos.donohoe.info</a> Repo: <a href="https:&#x2F;&#x2F;github.com&#x2F;dougdonohoe&#x2F;ddphotos" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;dougdonohoe&#x2F;ddphotos</a>

[Other] Show HN: Local-first firmware analyzer using WebAssembly Hi HN,<p>I just wanted to share what I have been working on for the past few months: A firmware analyzer for embedded Linux systems that helps uncovering security issues running entirely in the browser.<p>This is a very early Alpha. It is going to be rough around the edges. But I think it provides quite a lot of value already.<p>So please go ahead and drop a firmware (only .tar rootfs archives for now) and try to break it :)

[Testing] Test your prompts, agents, and RAGs. AI Red teaming, pentesting, and vulnerability scanning for LLMs. Compare performance of GPT, Claude, Gemini, Llama, and more. Simple declarative configs with command line and CI/CD integration.

[Other] OpenTelemetry for Rust Developers

[Other] Removing recursion via explicit callstack simulation

[Other] Claude Code, Claude Cowork and Codex #5

[Other] Show HN: I Was Here – Draw on street view, others can find your drawings Hey HN, I made a site where you can draw on street-level panoramas. Your drawings persist and other people can see them in real time.<p>Strokes get projected onto the 3D panorama so they wrap around buildings and follow the geometry, not just a flat overlay. Uses WebGL2 for rendering, Mapillary for the street imagery.<p>The idea is for it to become a global canvas, anyone can leave a mark anywhere and others stumble onto it.

[Other] The Cost of 'Lightweight' Frameworks: From Tauri to Native Rust

[Other] Oracle is building yesterday's data centers with tomorrow's debt

[Other] Code Review for Claude Code

[Other] Show HN: The Mog Programming Language Hi, Ted here, creator of Mog.<p>- Mog is a statically typed, compiled, embedded language (think statically typed Lua) designed to be written by LLMs -- the full spec fits in 3,200 tokens. - An AI agent writes a Mog program, compiles it, and dynamically loads it as a plugin, script, or hook. - The host controls exactly which functions a Mog program can call (capability-based permissions), so permissions propagate from agent to agent-written code. - Compiled to native code for low-latency plugin execution -- no interpreter overhead, no JIT, no process startup cost. - The compiler is written in safe Rust so the entire toolchain can be audited for security. Even without a full security audit, Mog is already useful for agents extending themselves with their own code. - MIT licensed, contributions welcome.<p>Motivations for Mog:<p>1. Syntax Only an AI Could Love: Mog is written for AIs to write, so the spec fits easily in context (~3200 tokens), and it&#x27;s intended to minimize foot-guns to lower the error rate when generating Mog code. This is why Mog has no operator precedence: non-associative operations have to use parentheses, e.g. (a + b) * c. It&#x27;s also why there&#x27;s no implicit type coercion, which I&#x27;ve found over the decades to be an annoying source of runtime bugs. There&#x27;s also less support in Mog for generics, and there&#x27;s absolutely no support for metaprogramming, macros, or syntactic abstraction.<p>When asking people to write code in a language, these restrictions could be onerous. But LLMs don&#x27;t care, and the less expressivity you trust them with, the better.<p>2. Capabilities-Based Permissionsl: There&#x27;s a paradox with existing security models for AI agents. If you give an agent like OpenClaw unfettered access to your data, that&#x27;s insecure and you&#x27;ll get pwned. But if you sandbox it, it can&#x27;t do most of what you want. Worse, if you run scripts the agent wrote, those scripts don&#x27;t inherit the permissions that constrain the agent&#x27;s own bash tool calls, which leads to pwnage and other chaos. And that&#x27;s not even assuming you run one of the many OpenClaw plugins with malware.<p>Mog tries to solve this by taking inspiration from embedded languages. It compiles all the way to machine code, ahead of time, but the compiler doesn&#x27;t output any dangerous code (at least it shouldn&#x27;t -- Mog is quite new, so that could still be buggy). This allows a host program, such as an AI agent, to generate Mog source code, compile it, and load it into itself using dlopen(), while maintaining security guarantees.<p>The main trick is that a Mog program on its own can&#x27;t do much. It has no direct access to syscalls, libc, or memory. It can basically call functions, do heap allocations (but only within the arena the host gives it), and return something. If the host wants the Mog program to be able to do I&#x2F;O, it has to supply the functions that the Mog program will call. A core invariant is that a Mog program should never be able to crash the host program, corrupt its state, or consume more resources than the host allows.<p>This allows the host to inspect the arguments to any potentially dangerous operation that the Mog program attempts, since it&#x27;s code that runs in the host. For example, a host agent could give a Mog program a function to run a bash command, then enforce its own session-level permissions on that command, even though the command was dynamically generated by a plugin that was written without prior knowledge of those permission settings.<p>(There are a couple other tricks that PL people might find interesting. One is that the host can limit the execution time of the guest program. It does this using cooperative interrupt polling, i.e. the compiler inserts runtime checks that check if the host has asked the guest to stop. This causes a roughly 10% drop in performance on extremely tight loops, which are the worst case. It could almost certainly be optimized.)<p>3. Self Modification Without Restart: When I try to modify my OpenClaw from my phone, I have to restart the whole agent. Mog fixes this: an agent can compile and run new plugins without interrupting a session, which makes it dynamically responsive to user feedback (e.g., you tell it to always ask you before deleting a file and without any interruption it compiles and loads the code to... actually do that).<p>Async support is built into the language, by adapting LLVM&#x27;s coroutine lowering to our Rust port of the QBE compiler, which is what Mog uses for compilation. The Mog host library can be slotted into an async event loop (tested with Bun), so Mog async calls get scheduled seamlessly by the agent&#x27;s event loop. Another trick is that the Mog program uses a stack inside the memory arena that the host provides for it to run in, rather than the system stack. The system tracks a guard page between the stack and heap. This design prevents stack overflow without runtime overhead.<p>Lots of work still needs to be done to make Mog a &quot;batteries-included&quot; experience like Python. Most of that work involves fleshing out a standard library to include things like JSON, CSV, Sqlite, and HTTP. One high-impact addition would be an `llm` library that allows the guest to make LLM calls through the agent, which should support multiple models and token budgeting, so the host could prevent the plugin from burning too many tokens.<p>I suspect we&#x27;ll also want to do more work to make the program lifecycle operations more ergonomic. And finally, there should be a more fully featured library for integrating a Mog host into an AI agent like OpenClaw or OpenAI&#x27;s Codex CLI.

[Other] Building a Procedural Hex Map with Wave Function Collapse

[Other] Show HN: DenchClaw – Local CRM on Top of OpenClaw Hi everyone, I am Kumar, co-founder of Dench (<a href="https:&#x2F;&#x2F;denchclaw.com" rel="nofollow">https:&#x2F;&#x2F;denchclaw.com</a>). We were part of YC S24, an agentic workflow company that previously worked with sales floors automating niche enterprise tasks such as outbound calling, legal intake, etc.<p>Building consumer &#x2F; power-user software always gave me more joy than FDEing into an enterprise. It did not give me joy to manually add AI tools to a cloud harness for every small new thing, at least not as much as completely local software that is open source and has all the powers of OpenClaw (I can now talk to my CRM on Telegram!).<p>A week ago, we launched Ironclaw, an Open Source OpenClaw CRM Framework (<a href="https:&#x2F;&#x2F;x.com&#x2F;garrytan&#x2F;status&#x2F;2023518514120937672?s=20" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;garrytan&#x2F;status&#x2F;2023518514120937672?s=20</a>) but people confused us with NearAI’s Ironclaw, so we changed our name to DenchClaw (<a href="https:&#x2F;&#x2F;denchclaw.com" rel="nofollow">https:&#x2F;&#x2F;denchclaw.com</a>).<p>OpenClaw today feels like early React: the primitive is incredibly powerful, but the patterns are still forming, and everyone is piecing together their own way to actually use it. What made React explode was the emergence of frameworks like Gatsby and Next.js that turned raw capability into something opinionated, repeatable, and easy to adopt.<p>That is how we think about DenchClaw. We are trying to make it one of the clearest, most practical, and most complete ways to use OpenClaw in the real world.<p>Demo: <a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=pfACTbc3Bh4#t=43" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=pfACTbc3Bh4#t=43</a><p><pre><code> npx denchclaw </code></pre> I use DenchClaw daily for almost everything I do. It also works as a coding agent like Cursor - DenchClaw built DenchClaw. I am addicted now that I can ask it, “hey in the companies table only show me the ones who have more than 5 employees” and it updates it live than me having to manually add a filter.<p>On Dench, everything sits in a file system, the table filters, views, column toggles, calendar&#x2F;gantt views, etc, so OpenClaw can directly work with it using Dench’s CRM skill.<p>The CRM is built on top of DuckDB, the smallest, most performant and at the same time also feature rich database we could find. Thank you DuckDB team!<p>It creates a new OpenClaw profile called “dench”, and opens a new OpenClaw Gateway… that means you can run all your usual openclaw commands by just prefixing every command with `openclaw --profile dench` . It will start your gateway on port 19001 range. You will be able to access the DenchClaw frontend at localhost:3100. Once you open it on Safari, just add it to your Dock to use it as a PWA.<p>Think of it as Cursor for your Mac (also works on Linux and Windows) which is based on OpenClaw. DenchClaw has a file tree view for you to use it as an elevated finder tool to do anything on your mac. I use it to create slides, do linkedin outreach using MY browser.<p>DenchClaw finds your Chrome Profile and copies it fully into its own, so you won’t have to log in into all your websites again. DenchClaw sees what you see, does what you do. It’s an everything app, that sits locally on your mac.<p>Just ask it “hey import my notion”, “hey import everything from my hubspot”, and it will literally go into your browser, export all objects and documents and put it in its own workspace that you can use.<p>We would love you all to break it, stress test its CRM capabilities, how it streams subagents for lead enrichment, hook it into your Apollo, Gmail, Notion and everything there is. Looking forward to comments&#x2F;feedback!

[Other] 169 production-ready skills & plugins for Claude Code, OpenAI Codex, and OpenClaw — engineering, marketing, product, compliance, C-level advisory, and more. Install via /plugin marketplace.

[Other] Show HN: VS Code Agent Kanban: Task Management for the AI-Assisted Developer Agent Kanban has 4 main features:<p>GitOps &amp; team friendly kanban board integration inside VS Code Structured plan &#x2F; todo &#x2F; implement via @kanban commands Leverages your existing agent harness rather than trying to bundle a built in one .md task format provides a permanent (editable) source of truth including considerations, decisions and actions, that is resistant to context rot

[Other] Show HN: Husky hook that blocks Git push until you do your pushups

[Other] Show HN: Satellite imagery object detection using text prompts I built a browser-based tool for detecting objects in satellite imagery using vision-language models (VLMs). You draw a polygon on the map and enter a text prompt such as &quot;swimming pools&quot;, &quot;oil tanks&quot;, or &quot;buses&quot;. The system scans the selected area tile-by-tile and returns detections projected back onto the map as GeoJSON.<p>Pipeline: select area and zoom level, split the region into mercantile tiles, run each tile with the prompt through a VLM, convert predicted bounding boxes to geographic coordinates (WGS84), and render the results back on the map.<p>It works reasonably well for distinct structures in a zero-shot setting. occluded objects are still better handled by specialized detectors like YOLO models.<p>There is a public demo and no login required. I am mainly interested in feedback on detection quality, performance tradeoffs between VLMs and specialized detectors, and potential real-world use cases.