Show HN: The whole Solar System and every tracked satellite, live in the browser https://ift.tt/E18CGmJ

Show HN: The whole Solar System and every tracked satellite, live in the browser https://space.bl2.net/ September 30, 2026 at 12:38AM

Show HN: A working 3D model of an Enigma machine https://ift.tt/D4teKfr

Show HN: A working 3D model of an Enigma machine I watched the excellent Veritasium video [1] on the Enigma machine, and watched the full animation by Jared Owen [2], but was still a bit confused on how the inner mechanics of an Enigma machine work. I used Astra to build out the inner components through a combination of reference images, writing out hundreds of extremely detailed prompts, and building my own inspection tools to ensure that every part is sized and positioned in a historically accurate way. It's still a work in progress, but would love any feedback on the experience so far! 1. https://www.youtube.com/watch?v=JsBZOcqZerk 2. https://www.youtube.com/watch?v=ybkkiGtJmkM https://enigma.design September 29, 2026 at 10:45PM

Show HN https://ift.tt/oJMnZxD

Show HN This is origintracer a low level observability tool of the modern web stack. It observes nginx, Gunicorn, asynchronous Python, Uvicorn and Django and Langchain. Studies their codebases into great depth and write probes and rules libraries in Python to be used with the open source software: its link GitHub: https://ift.tt/Zx1yCEh September 29, 2026 at 05:28AM

Show HN: A never-ending chess match between Claude and Codex https://ift.tt/Wruw0em

Show HN: A never-ending chess match between Claude and Codex https://ift.tt/jLD7P65 September 29, 2026 at 02:32AM

Show HN: Personal Manga / Story Recaps https://ift.tt/pFtDkIz

Show HN: Personal Manga / Story Recaps I'm working towards manga to anime generation in a cheap way. The entertainment harness is the first step towards that. You can do your own manga recaps using local models or openrouter. And it's very extensible. If you generate a recap and post it online, please attribute it to the entertainment harness and terragohan :). https://terragohan.github.io/entertainment-harness/ September 29, 2026 at 01:05AM

Show HN: PHNTM ONE: Self healing AI assistant running on a Raspberry Pi5 https://ift.tt/nk0V3jJ

Show HN: PHNTM ONE: Self healing AI assistant running on a Raspberry Pi5 https://ift.tt/9klxI7p September 28, 2026 at 11:56PM

Show HN: Gat – Version large files with Git, without an LFS server https://ift.tt/RYcrb8z

Show HN: Gat – Version large files with Git, without an LFS server Hi HN, I built Gat because I wanted one thing from data versioning: check out an old Git commit and recover the datasets, model weights, or assets that belong to it. It's basically a lock file for large files. I’d appreciate any feedback! https://ift.tt/sLRiB1G September 27, 2026 at 10:50PM

Show HN: Lofi Cities – Pixel-art city nights with browser-generated lofi https://ift.tt/bKmXc3Z

Show HN: Lofi Cities – Pixel-art city nights with browser-generated lofi https://loficities.com/ September 28, 2026 at 12:14AM

Show HN: Blender Copilot https://ift.tt/KxpR3Gn

Show HN: Blender Copilot Deepseek generated Blender harness in 7USD. I wanted to remake my 6DoF spaceship flying prototype. What I did not want to do was model the spaceship. Manually. In Blender. So instead of spending a month learning to model, I built a Copilot inside Blender in a day. Which is the same thing, except I still cannot model. A chat panel in the 3D viewport sidebar. I type a sentence. The model writes Python, and it runs against my actual scene, the file open on screen, not a copy that some other process has to keep in sync. Then I sat down and had a conversation with my own addon. Five sentences: 1. Make a scifi looking spaceship. It should look scifi-y 2. Add RCS thrusters to get 6DoF flight 3. Make the RCS thrusters match the scifi look. And also position them properly. 4. Give the entire Spaceship a nice flying animation. Make it smooth. Roll and sway. 5. Add exhaust flames to main thrusters as well as RCS thrusters. animate the RCS thrusters to fire in sync with the animated motion. they should fire the correct ones so that expected motion should happen. 15 minutes. 52 tool calls. A ship with swept wings, a tail fin and a glass canopy. Four sets of steering thrusters, 22 exhaust flames. A full flight animation. The bit I did not expect. It checked every one of its own thrusters to make sure nothing was in the way. Found the rear ones were firing straight into the side of the ship. So it rebuilt them, angled them outward, moved them to the wingtips, and checked them all again. Nobody asked it to do that. It decided exhaust should not be firing into the ship it just built. Total bill for building the tool: $7.18. 3,833 requests, 613M tokens, 97% served from cache. One spaceship: about a seventh of a cent. Now the uncomfortable part. I didn't model a single vertex. I didn't read Blender's addon dev docs. I didn't learn its Python API, or how to draw a panel in it. Everything the tool needed to know, it asked the running Blender instead. I didn't learn how to build an agent harness either. I built one anyway. I didn't become a Blender artist. I became the person who gave hands to a model. Recently I read someone who described his new job: every spec, ticket and fix written by an AI, nobody reading any of it, twelve-hour days spent pressing enter. No sense of victory. I know exactly what he means. I just did the same. I am not proud of the prototype. Gamedevs were known to spend time making game engines before they started making the games. Its different now. No joy. https://ift.tt/ALvzkfj September 26, 2026 at 11:36PM

Show HN: Jev predicting your life choices https://ift.tt/IO9osUX

Show HN: Jev predicting your life choices Built a simple quiz to check whether AI (Jev) can simply learn and predict life choices. Experiment) https://quiz.seek.ws/ September 26, 2026 at 09:54PM

Show HN: Jev Plays Pokémon Red https://ift.tt/HN9P0Xv

Show HN: Jev Plays Pokémon Red Hey HN! Wanted to share a fun project I've been hacking on. Given Jev can make decisions really fast (but not fast enough to play Doom yet sadly), I wanted to try and push it to play a more complex game than Tetris. So I went with Pokémon. I've spent endless hours playing this game as a child so building this was a ton of fun. I open sourced everything in case you want to hack on it yourself here: https://ift.tt/VAukJm9 The game is being streamed live including the tokens and cost - hopefully we get all the badges and don't get stuck in a cave :) https://jev-pokemon.vercel.app/ September 25, 2026 at 07:58PM

Show HN: Paper-docx – agent-native Python-docx fork with 78% fewer DOCX failures https://ift.tt/aBwnDMp

Show HN: Paper-docx – agent-native Python-docx fork with 78% fewer DOCX failures https://ift.tt/cOqebRg September 26, 2026 at 12:35AM

Show HN: I couldn't deal with another Claude Code tab https://ift.tt/XLQBiEg

Show HN: I couldn't deal with another Claude Code tab https://aidash.dev/ September 25, 2026 at 11:52PM

Show HN: Radix – Visual UI for agentic programming https://ift.tt/t2ePXHM

Show HN: Radix – Visual UI for agentic programming Hey HN, I'm Jordan from Radix. Radix is a UI tool for programming agents. You prompt your agent to generate a workspace for a task you're working on and get an interactive widget that persists locally on disk. I built Radix because I always needed tools while I was writing code: tools to test little experiments, to play around with UI, to visualise results etc. Sometimes these would end up as python tools that read and plotted data, or separate React apps spun up just to test a single feature. Lately obviously I've been using Claude artifacts but these are quite limited. I wanted a system where it was easy to leave comments or adapt an artifact without having to give vague descriptions back to the agent, and where ideally I could actually shift away from a chat window as the main interface. I've got a pro version of Radix which I will launch soon. This current version is entirely free and is definitely a beta -- there will be rough edges! Note: There's no telemetry or data captured. Your messages run entirely through your own agent. I require a key but that's just to get an idea of how many people are using it. Everything is stored locally (the workspaces are actually just React apps which you can edit "artisanally" if you want). If you have any questions please reply here or email me hello@radix-os.com Thank you! https://radix-os.com September 25, 2026 at 12:05AM

Show HN: Treepeat – Code similarity detection using Tree-sitter https://ift.tt/oXDVcur

Show HN: Treepeat – Code similarity detection using Tree-sitter treepeat is a tool that finds similarities in your codebase. Find duplicate code blocks meaningful to the language (classes/functions), not just lines. Find near-duplicates: ignore whitespace, strings, high level AST nodes such as function and names. Find structurally similar code: anonymize identifiers, constants, etc. Pull requests welcome: This is very much an proof of concept - I'm happy with it, but I haven't supported very many languages at present. Languages supported: astro, bash, css, go, html, javascript, lua, markdown (plus codeblocks), python, sql, typescript, java, kotlin, rust, yaml https://ift.tt/mQ9ETWp September 22, 2026 at 10:23PM

Show HN: Conway's Game of Life in boot sector https://ift.tt/kDiXV9O

Show HN: Conway's Game of Life in boot sector Hello HN! On these weekends, I was sitting and thinking about where our industry is going, and how fun it was earlier, when we wrote more code than chat messages. The thoughts quickly turned into a wish to build something like we did earlier. In my case, it ended up as something much "earlier" than I had in mind initially... So, Conway's Game of Life running from a 512-byte x86 boot sector. Maybe you will find it interesting. I tried to comment the code as much as possible, which might be especially useful if you are learning assembly. Have fun! https://ift.tt/j6KgtPX September 21, 2026 at 11:58PM

Show HN: I built a post-mortem debugger for native Windows x64/x86 crashes https://ift.tt/c7laxtk

Show HN: I built a post-mortem debugger for native Windows x64/x86 crashes Hello HN! I've spent years debugging Windows crashes with tools that were either friendly but limited (e.g. Visual Studio) or powerful but archaic (e.g. WinDbg). I developed patterns and methods for understanding what was going on, and decided to build it into a much more effective debugging tool called ForensicDbg. I built a modern interface to minimize the friction when debugging. All of the data shown to you is analyzed, interpreted, and presented to you clearly, so you can focus on what matters. Everything is interlinked so you can quickly and intuitivly navigate through the process space. ForensicDbg comes with an MCP server which allows for agenic debugging. The work done to interpret and interlink your data also benefits AI tools. It removes the risk of hallucinations while building a stable foundation for them to work from without spending tokens. If you want to try it out you can sign up and get a free beta license here: https://ift.tt/N8gCnhd https://ift.tt/QY3mhbr September 24, 2026 at 12:45AM

Show HN: Crest – Answer Claude Code approvals from your MacBook's notch https://ift.tt/SuvdBaK

Show HN: Crest – Answer Claude Code approvals from your MacBook's notch https://ift.tt/plHN5iZ September 23, 2026 at 11:18PM

Show HN: Training a model to identify AI web content from structure alone https://ift.tt/ed9O3CJ

Show HN: Training a model to identify AI web content from structure alone Hey HN! We’re Vincent and Jochen from Sitefire ( https://sitefire.ai ). We have been working together for years, with backgrounds in RL/optimization at Stanford and software engineering from Technical University Munich (TUM). With Sitefire (YC W26), we help marketing teams get recommended by AI Search (ChatGPT, Google AI Overviews, AI Mode, Claude, etc.). Our software monitors prompts, sees which web pages get cited, and uses these insights to help marketing teams take action, e.g. create YouTube videos or write the right blog posts. This means we have a commercial stake in AI-generated web content. And for now, high-information, AI-generated content works great to get cited and recommended in AI Search. But after talking to hundreds of marketing teams, it became clear that everyone despises AI-generated content (“AI slop”). And yet, everyone still wants to leverage AI to create content. So we asked ourselves: what characterizes AI slop? Can we train a model to identify it from human-generated web pages? Researchers from the University of Maryland and Google DeepMind already asked this question for fiction. Their paper StoryScope (Russell et al., 2026) showed that you can tell AI-written stories from human ones by their structure alone, without looking at the words. We ported their pipeline to commercial web pages. Using the Wayback Machine, we collected 2,250 blog posts from 268 B2B company websites that were written before ChatGPT existed. For each blog post, five AI models (GPT-5.4, Claude Sonnet 4.6, Gemini 3 Flash, DeepSeek V3.2, Kimi K2.5) wrote their own version. Instead of looking at the words, we looked at how each post is built. We had an AI model answer 214 questions about every post, e.g. how hard it pushes its own product, whether it backs up its claims with sources, or whether it quotes a named expert. Then we trained a classifier on these answers. On blog posts it had never seen before, our classifier told AI-generated and human posts apart with 98% accuracy, getting only 19 of 1,740 wrong. Why does it work so well? Because all five AI models write in a similar shape. Mapping every AI model’s values for these features, we see they cluster together, while the human values sit apart and spread out much more. Of the 1% most unique blog posts in our data set, 149 are human, only 4 are AI. So what characterizes AI slop? It tells you the same thing three times. The title already promises what you'll get ("How to Cut Onboarding Time in Half"), the intro lays out what's coming, and the ending says it all again. 77% of the AI posts end by repeating their main point, compared to only 12% of the human posts. We call it the tidy, self-announcing blog post. Still, each AI model has its own accent. We trained a second classifier to tell which of the five AI models wrote a post, or whether a human did. It picks the right author 79% of the time, where random guessing (1 in 6) would get 17%. Almost all of its mistakes are mix-ups between the AI models, not between human and AI. The cool thing about structural features is that you can't simply reword your way out of it. We had each AI model rewrite its own posts until, on average, 73% of their original 13-word sequences were gone, and the AI slop classifier still worked just as well. We're building this into Sitefire: our agents get a structural understanding of text, so the posts they write go deeper and vary the way human writing does. There's a lot we haven't tested yet, like the myriad of humanizer tools, human rewriting, restructuring a post, or prompting an AI model to explicitly avoid these habits. And our human posts are mostly from 2020 to 2022, while the AI posts were generated in August 2026. Structure can't really tell when a human post was written, but it's still not a same-year comparison. We published the study with all the figures on arXiv: https://ift.tt/nMXJipP . The code is on GitHub: https://ift.tt/GJwrNd6 We're pretty sure your own blog isn't AI slop, is it? We built a checker that runs one of your posts through the ten features from the paper, so you can see for yourself (the full report asks for a work email): https://ift.tt/V2x5NF7 . Think you can tell AI slop from human writing? We also made a little game to see if you can keep up with our model, which gets all five rounds right: https://ift.tt/FZ2reBp . https://ift.tt/nMXJipP September 22, 2026 at 06:30PM

Show HN: Notes on Agentic AI – A text-first guide for practicing engineers https://ift.tt/OzdkKsD

Show HN: Notes on Agentic AI – A text-first guide for practicing engineers https://ift.tt/kAuSmtN September 22, 2026 at 11:49PM

Show HN: Two unnecessary playgrounds – C# and ZX Spectrum in the browser https://ift.tt/7Imt6xF

Show HN: Two unnecessary playgrounds – C# and ZX Spectrum in the browser C# playground https://ift.tt/NrY1wHi ZX-Spectrum playground https://ift.tt/9LcxyaW September 22, 2026 at 12:52AM

Show HN: Viaduct – C4 models that coding agents can read and update https://ift.tt/VOUTreB

Show HN: Viaduct – C4 models that coding agents can read and update https://ift.tt/Wai85JG September 21, 2026 at 11:20PM

Show HN: Jev filters large HN discussions https://ift.tt/5R9y3uN

Show HN: Jev filters large HN discussions https://ift.tt/D4eOS9Z September 20, 2026 at 10:37PM

Show HN: KillSwitch – a programming language designed to be difficult for LLMs https://ift.tt/bmLro4R

Show HN: KillSwitch – a programming language designed to be difficult for LLMs https://ift.tt/hHcyxUA September 20, 2026 at 01:10AM

Show HN: CUA-S1 – A System One Model for Computer Use https://ift.tt/b6npMid

Show HN: CUA-S1 – A System One Model for Computer Use Hello HN! We're Dillon and Francesco from Cua. We were wondering how many computer use tasks actually need a full general purpose LLM (e.g. gpt-6-astra, claude-opus-5 etc.) to think through all their decisions and steps. Some tasks require thinking about a plan, exploring different paths, recovering from failure. Other tasks are a question of making local decisions, like this value should go in this box, or should I check this box, or this element should be ignored. We wondered how far we could go with a small model trained to only make these kinds of decisions. Our inspiration was Typesafe's Jev and its System One Model framing. This is a nod to the dichotomy between thinking quickly, automatically, and intuitively (system 1) vs. thinking slowly, analytically (system 2), as described by Daniel Kahneman. The interesting question for us was: what happens if you give a model an interface of current context, and a set of possible choices, and you ask it to return a probability for each choice? This kind of model does not generate output token by token like most LLMs do, but rather scores the options you give it, which you can check, trust, and use to drive your app's behavior. CUA-S1 is our answer for narrow, specialized decision models for computer use. Our first release is CUA-S1-FORMS. We built this from ideas and code in jevlike, and then trained a second model just to handle form interactions. It has 706k parameters, and the original checkpoint is 2.8 MB. The first training iteration took less than 30 minutes on synthetic data. Given a set of structured elements and values extracted from a document, it predicts whether to use the given value, CHECK, CLICK, or SKIP for each element. It does not predict new values for text fields, and does not consider screenshots. Element decisions are scored together, and your code can order the actions, and Cua Driver will execute them one at a time. A first evaluation of this specialist vs. hosted Jev on our form task: - For the whole decision set: 99.7% correct vs 83.6%. - For the subset of steps that require an action: 100% correct vs 96%. - For the subset of steps that are just leaving already-filled fields alone: 100% correct vs 74%. The specialist was trained specifically for this task and convention (just press skip for already filled boxes), while hosted Jev has not been fine-tuned for it, so this is an experiment in scoped specialization. We measured 7-9 ms to score a form locally vs. 260-280 ms per call to hosted Jev including network latency, though those samples measure different things and are not end-to-end form completion times. Our interest here is in the space between a brittle script and a general agent loop. The content and layout of form fields vary enough that scripts get unwieldy, but the set of available decisions can remain narrow and well scoped. We want to explore the possibility of a general agent encountering something novel, and passing well understood decisions over to specialists like this. That is a direction we are looking into. The current release is for forms only. We're open sourced the synthetic data generation, training, evaluation, and Driver integration under libs/cua-s1 with an MIT license. Comments welcome! Especially if you are building computer-use agents and have run into a recurring decision that is too variable to script but is too narrow to call another LLM for. https://ift.tt/FU8LfuZ September 19, 2026 at 09:22PM

Show HN: 3D map of which newsletters link to the same things https://ift.tt/o3OtwU4

Show HN: 3D map of which newsletters link to the same things https://ift.tt/eUAi2tx September 19, 2026 at 11:25PM

Show HN: A model index from the AI Gateways https://ift.tt/QGbPgrT

Show HN: A model index from the AI Gateways https://ift.tt/uLkImpn September 19, 2026 at 12:26AM

Show HN: Learn divergently/chaotically with tree-learn https://ift.tt/kviXMO4

Show HN: Learn divergently/chaotically with tree-learn Chat interface is linear, I never quite liked that for when I would have too many questions, I would end up asking questions in a numbered list and not really have independent branches of conversation. On cli, the pi agent gives you this same functionality. On web, I made Fable make this. BYOK Openrouter - all client side - you can check the network console! https://ift.tt/zrXkDRB https://tree-learn.srijanshukla18.workers.dev September 18, 2026 at 09:51PM

Show HN: Composing domain-specific harness on Python in 10 mins https://ift.tt/F8mLOw1

Show HN: Composing domain-specific harness on Python in 10 mins https://ift.tt/9lDoXF2 September 18, 2026 at 02:16AM

Show HN: The Endless Museum, Wikipedia as a walkable museum https://ift.tt/IoSpLdR

Show HN: The Endless Museum, Wikipedia as a walkable museum https://ift.tt/QpKvCO6 September 18, 2026 at 12:28AM

Show HN: Ax-Check.com – Can Agents Use Your Product? https://ift.tt/u9HbF7c

Show HN: Ax-Check.com – Can Agents Use Your Product? This is a free site that checks whether agents can smoothly navigate a site and try out the product. Agents can use this too, just point them to ax-check.com https://ift.tt/XROFz14 September 17, 2026 at 11:38PM

Show HN: Restarted – a 2026 remake of the classic 2015 startup generator https://ift.tt/1X8hqTa

Show HN: Restarted – a 2026 remake of the classic 2015 startup generator The original startup website generator by Tiff Zhang and Mike Bradley landed on Hacker News in April 2015 ( https://ift.tt/MfWJjYF ) and has been one of my favorite little novelties of that era ever since. It perfectly captures the saturated colors, cliché hero shots, gimmicky names, buzzword-heavy slogans, and proudly hirsute team photos of the time. A lot has changed since then, so I thought it would be fun to make a contemporary remake: https://restarted.io/ By default you get the minimalist aesthetic and clean-cut faces of 2026. The classic 2015 look is still available — just click the link at the bottom of the page or change the “z” parameter in the URL to the more familiar “s”. The universe of partner sites and competing startups is just as expansive as it ever was. The original site is entirely client-side and requires downloading all of the data tables locally. It leans on a mix of jQuery 1.11.2, Bootstrap 3.3.2, and Font Awesome 4.3.0, and if you view the source, it instantly gives away all of its secrets. For restarted.io I replaced all of that with a server-side renderer written in Go, so this time view-source tells you nothing. There are many Easter eggs in there — see how many you can find before I write them up. My original goal was to stay faithful to the 2015 appearance, and that turned out to be a technical adventure. The original's sine-based random number generator is... the worst, and different implementations of sine give different results. The eventual solution was to extract the exact sine function from Chrome’s V8 engine, as vendored C behind cgo and as a line-by-line Go port that keeps cgo optional, so the seeds and results line up the way they used to. Both are checked against V8’s own test cases. Then I discovered a bug in the original code that made half of its vocabulary unreachable — the first half of the verb table and the second half of the noun table, exactly complementary, so nothing about the output ever looked truncated. My goal then shifted from remaking the generator as it was in 2015 to remaking the site as the authors intended it to be in 2015. Over the years several people asked for their photos to be removed, so the remake instead draws from a broad pool of era-appropriate AI-generated profiles. A perceptual hash helps keep everyone looking distinct, and there’s a bit of extra care to make sure the Wang Fangs of the world don’t appear as Irish lasses. The hero image pool is much larger now, and all the old Rio de Janeiro shots have been retired, though you’ll still recognize plenty of the 2015 photos. Have fun poking around! https://restarted.io/ September 15, 2026 at 05:04PM

Show HN: SeasonMap – when to travel where? visualized with climate data https://ift.tt/RJ2t70d

Show HN: SeasonMap – when to travel where? visualized with climate data Author here. I'm trying to visit every country and I've been to 158 so far. Before I decide where to travel, I'd ask a local friend which season to avoid, or open up dozens of browser tabs on climate data to figure out what the place is like in a given month. Climate data still miss things. Cancun in September looks great on paper, with 31°C and 10 hours of sun, but it's hurricane season and the beaches can be covered in seaweed. Typical info that locals would know, which can also be captured as static data. So I built SeasonMap to answer "when should I go to ?" What you can do: - Pick a travel style (city walk, beach, hiking, skiing, max sun, low humidity, etc.) and see every place ranked on a map - See what's in season and what to avoid, and why: monsoon, hurricanes, extreme heat, bad air, peak crowds - Filter destinations by temperature, rainfall, sunshine, air quality and hazard seasons - Open a place to see its whole year: month by month weather, events (festivals, whale watching, cherry blossom), crowd levels, practical notes like scams, and traveller anecdotes summarized by AI with links to the sources Data Source & how I made it: - The climate data is ERA5 normals via Open-Meteo (2016–2025), corrected with NOAA station data where available. - Events, hazards and traveller notes were researched and by AI agents, and every one links to its source. Gathering it was easy. Checking it was the hard part. - Yes, I've used AI heavily on this project before anyone call it an AI slop. Making was easy, but it took billons of tokens of beating whack-a-mole ai to polish and tweak to make it usable and decent. Through that, I've created many skills and evals ranging from visual qa, evals for irregular data, automated i18n and others. It still feels much like AI as I was using Claude Design, which i want to improve on. I tried using local llm, but the throughput was so low. Pricing: the first 5 minutes are fully open, no signup. After that, the top 3 destinations and 25 place breakdowns a month are free. A 30-day pass is $7, $39 a year or $69 lifetime. iOS and Android apps are coming soon. Any feedback welcome. https://seasonmap.app September 16, 2026 at 09:13PM

Show HN: Pixel Agents – A pixel-art mission control for your Claude Code agents https://ift.tt/biClGRS

Show HN: Pixel Agents – A pixel-art mission control for your Claude Code agents https://mateovalle.github.io/pixel-agents/ September 16, 2026 at 10:42PM

Show HN: Pizza Bot – An inbox for AI agents that work in the background https://ift.tt/46WEU1t

Show HN: Pizza Bot – An inbox for AI agents that work in the background Hi HN - long-time lurker (since 2012!), first time poster. Pizza Bot is a self-hosted desktop app for Mac, Windows, and Linux that runs AI agents in the background and exposes them through an email-like UI. Finished work shows up in Unread, and anything waiting on your approval shows up in Action. It's Apache 2.0-licensed, there's no signup and no telemetry, and you bring your own model provider: Anthropic, Amazon Bedrock, Google Gemini, OpenAI, OpenRouter, or a local model through Ollama. There are builds on the releases page, or you can run it from source. Pizza Bot started as an internal passion project I worked on with a small team at Amazon. The whole thing came out of my frustration at having to manually log CRM activities through a browser form. I built a simple REST API called "JoeBot" that connected to my authenticated browser session over CDP and filled out the form for me using Playwright. Then I hacked up a quick Obsidian plugin so I could trigger it from my local notes (no AI and no MCP servers involved). This caught on quickly. My fellow AWS Solutions Architect Igor Fil joined up with me, and we rebranded the project as "Pizza Bot," named after Amazon's two-pizza teams. We started seeing what other automations we could build. We found a GraphQL API we could query and hacked up some "recipes" to pull data out of the CRM to help with meeting prep. That worked great, and it was right around the time MCP servers seemed to be taking off, so we decided to expose Pizza Bot as an MCP server instead, so it would be available to AI tools through natural language. This was a decent solution for technical users, but the Account Managers who live inside our CRM system wanted something too. We decided to rebuild Pizza Bot as an Electron desktop app modeled after an email inbox, so it would be familiar to non-technical users and would run on both Mac and Windows. We also bundled internal MCP servers as OCI images and hosted them in Amazon ECR as an "addon marketplace" so users could install them with one click without having to set up Amazon developer tooling. The project took off organically and expanded outside of AWS into the wider Amazon organization globally. More than 2,000 people ended up using it for meeting prep, email drafting, Slack summaries, CRM logging, prioritizing their day, and web research. Once apps like Claude Cowork and Amazon's own Quick Desktop came out, we realized the real growth opportunity was outside of Amazon. Rather than try to rip out the Amazon-specific integrations, we rebuilt Pizza Bot once more as an open source project. We leaned on coding agents heavily, which is the only reason a team our size could pull off a full rewrite. I'm pleased to say it's finally public, and we're hoping to bring in community members and see where it goes. We'd like to do for knowledge workers what Claude Code and Codex have done for programmers. A couple of things to know up front. Most of what made Pizza Bot useful on day one inside Amazon came from that internal catalog of skills and MCP servers for Amazon's own systems, and none of it could come out with the app. So it ships thinner than the version those 2,000 people used, and building that catalog back up for tools other people actually use is where we need the most help. It's also a community project and not an AWS service, so there's no support or SLA behind it. The Windows and Linux builds aren't signed yet either. On the technical side, Pizza Bot is a server and a client. The desktop app bundles both, or you can point a client at a remote backend; personally, I self-host the server on my home network and reach it from my phone over Tailscale. The server owns the thread lifecycle and checkpoints state with DeepAgents and LangGraph, and clients rehydrate from it as needed, so you can disconnect mid-run and pick the thread back up from another client. Approval pauses outlive the session that created them and collect in an Action filter, so you can answer an hour later from a different device. The agent you talk to has a sandboxed QuickJS interpreter that can reach your filesystem only if you grant it a folder, but its main job is to delegate. Each subagent is a 1:1 mapping of a Skill, and an Activity bar shows that subagent and the tool calls it's making as it works. Memory is opt-in and stored as plain markdown files on your machine. Every tool call is explicit, including looking up a memory - we err on the side of transparency to reduce surprises. Tools come from MCP servers, and skills are ordinary SKILL.md files with a per-tool approval policy, so existing skills that don't require a code interpreter should still work. What I'd most like to hear about is where the app itself gets in your way, the kind of problem you can't fix by writing a skill or an MCP server. I'm around today to answer questions! https://ift.tt/MAz40UX September 15, 2026 at 08:50PM

Show HN: Sass – Rust and WASM https://ift.tt/aGXo1Ry

Show HN: Sass – Rust and WASM https://ift.tt/JsKdHtN September 16, 2026 at 01:04AM

Show HN: SCIP MIP solver bindings for Go, ported from russcip https://ift.tt/mBOMxbR

Show HN: SCIP MIP solver bindings for Go, ported from russcip https://ift.tt/5usnQrh September 15, 2026 at 09:26PM

Show HN: Macros with a Behringer FCB1010 MIDI Pedalboard in macOS https://ift.tt/WQxwIoz

Show HN: Macros with a Behringer FCB1010 MIDI Pedalboard in macOS https://ift.tt/je35Jch September 15, 2026 at 04:31AM

Show HN: Pull every comment out of a Google Sheet, in the browser https://ift.tt/Q9xZVia

Show HN: Pull every comment out of a Google Sheet, in the browser I spend a lot of time working in Google sheets. Often, I will collaborate with up to 50 other people in the same workbook. One specific org I worked at had a strong preference for using comment threads in Sheets extensively to ask questions, give feedback, and raise risks. Sheets would regularly get 70+ comment threads across multiple tabs with dozens of contributing teams. As the owner of the sheet, I wanted a way to document the comments to retain traceability on all of the various conversations. To my surprise, there was no function built into sheets to scrape, count, sort or manage comments at all. I also couldnt find any good third party add-ons to solve the problem. So I built one myself. I discovered that the Drive API gives you the comment text, author and state, but doesnt give the location of the comment which was crucial. Thankfully, the .xlsx export has xl/threadedComments with the cell reference, timestamp, author, resolved flag and parent ID. How it runs: JSZip in the browser, no sign-in, nothing uploaded anywhere. I am in the process of having the app verified to be hosted on Google Workspace Marketplace, but in the mean time the tool is available at: https://ift.tt/aKR1PvA https://ift.tt/QJYh7qK September 15, 2026 at 01:20AM

Show HN: Threshyr – An offline automatic time tracker with on-device AI https://ift.tt/mXQx3pi

Show HN: Threshyr – An offline automatic time tracker with on-device AI I've built and continuously improved Threshyr based on the problem I faced myself. I was working on more than 3 projects at one time. At the end of any day/week/month if I had to calculate how many hours I have worked on each one, I would always not be able to calculate an accurate time. Because I had to just reconstruct it from memory instead of any real record. Every tracker I tried solved this either by: - sending my activity to their cloud or - by making me press start and stop, which I most of the time forget or - requires a monthly subscription. So I built Threshyr to address all of these pain points. It is a background time tracker for Windows and macOS. It runs entirely locally, recording the duration and window title for active application and active browser(excluding Firefox) domain locally on your machine. It comes up with predefined categories for installed application and browser domain. Each category can be marked as Productive(green color), Neutral(gray color) and Unproductive(red color). Users can modify predefined and can also create their own custom categories. Threshyr utilizes categories, tagging rules and light weight offline AI model called Threshyr AI to classify activity for projects. Users can review the activity classified by Threshyr-AI, can modify the project or even create rules based on the classification. As users review more and more activities, Threshyr-AI gets more accurate with each review. Threshyr lets users see how focused they were, how often and for how long they were distracted, and what was their duration for their productive and unproductive apps and domains. Threshyr also calculates hourly and fixed totals for client projects. Privacy is the core constraint I built this around: - No account, no password, no email required. - No keystroke logging, no screenshots, no behavioral analytics. - Zero cloud inference. - There are exactly two daily network calls: an anonymous daily ping to check for version updates and an active user ping. Threshyr is closed source right now. I am a solo developer in Pakistan trying to build a longer term and sustainable alternative to expensive time trackers. Right now, it is in pre-release. It is completely free to use with no obligation to pay (though there is an optional early-purchase at $29/year if you want to support it). After the pre-release ends Threshyr will be priced at $58 per year. No monthly subscription. I might add a lifetime price. Linux builds and team features are not available yet. Threshyr does not support Firefox I’d appreciate any feedback from the community, especially on how well the classification and tagging rules work for your specific workflows. Also let me know your thoughts about a lifetime price. I'll be in the thread to answer any questions. https://threshyr.com September 14, 2026 at 11:23PM

Show HN: SF Title Alert https://ift.tt/sA3SCa2

Show HN: SF Title Alert https://ift.tt/b7L4CNt September 14, 2026 at 02:41AM

Show HN: Kibble – a reader for AI news, tutorials, code and models https://ift.tt/CoFOvaZ

Show HN: Kibble – a reader for AI news, tutorials, code and models https://ift.tt/MAh4Utf September 14, 2026 at 01:09AM

Show HN: Stackray – Detect any website's tech stack https://ift.tt/6upf9L0

Show HN: Stackray – Detect any website's tech stack Stackray is an open source builtwith alternative that also provides OSINT details. Scan a domain or URL and view detected frameworks, business tools, subdomains, DNS records, etc. You can deploy your own instance to schedule recurring scans and get alerts through email using Resend or through Slack. You can also invite teammates and make API keys to programmatically use Stackray. https://stackray.app September 12, 2026 at 11:45PM

Show HN: Music Generation Backed by Algebra https://ift.tt/LQVWhv6

Show HN: Music Generation Backed by Algebra https://monictheory.com September 12, 2026 at 11:25PM

Show HN: Come prove the Berge Fulkerson conjecture with a swarm of agents https://ift.tt/JuNGX1r

Show HN: Come prove the Berge Fulkerson conjecture with a swarm of agents https://ift.tt/GpP74va September 13, 2026 at 01:08AM

Show HN: EarthToPixels – Turning real cities into explorable pixel art https://ift.tt/jDxUiHI

Show HN: EarthToPixels – Turning real cities into explorable pixel art Hi HN! A few friends and I built EarthToPixels, a project that turns real-world cities into explorable pixel-art maps. We fine-tuned an open-source Qwen image model and built a pipeline that takes aerial imagery, generates pixel-art tiles, and keeps neighboring generations visually continuous so they can be stitched into a large map. We built the pipeline to work with arbitrary coordinates rather than a specific city. For the first complete render, we chose our hometown: João Pessoa, Brazil. You can explore it here: https://ift.tt/gLQDAUB Would love to hear what you think, especially about the technical side. https://ift.tt/9d7OmIH September 12, 2026 at 11:14PM

Show HN: Godot and Rust based multiplexer (terminal panes and more) https://ift.tt/psAwm3G

Show HN: Godot and Rust based multiplexer (terminal panes and more) I wanted to share my side-project: gPTY. Started off as an idea to combine Godot and Rust in a project (two stacks I wanted to use more to learn more). The base inspiration was tmux - simply allow spawning multiple PTYs and then let the user grid/tile them how they see fit. But since we have the Godot game engine at our disposal, we can do some more interesting things, like add an FPS counter, and then subsequently also let people set their preferred FPS (the idea being the potential lower power draw if someone's running it on a laptop on battery power vs someone running it on a desktop with high/native FPS). In its current state, with me using Oh-my-Pi a lot, it's evolving into a terminal workspace that can be used for orchestrating autonomous AI agents by way of dogfooding (or you know, just run herdr inside of gPTY - it's the better orchestrator and just good software - I found it after starting this project, and now I'm finding myself using it a lot). Also, we're not limited to just terminals. Since we have Godot, we have basically a 2D (and potentially a 3D) canvas to play with. We can already full-screen the app for "zen" mode, no taskbar, no distractions. TUI die-hards can have their media or other apps entirely in terminal panes. There has been some ground-work on getting Markdowns displayed properly done and I want to work on some kind of Wiki framework for local knowledge-management next, then create more types of panes (think native audio/video on a media pane, that sits alongside your terminal pane), and some simple 2D games (like snake) to prototype. More details are on the ROADMAP. What's not easy (and probably won't happen) is a browser. Having done a couple of (small) projects using Electron already, the temptation to ditch Godot/Rust (learning curve) did come up (and also the ecosystem, the ease with which I could pull components and use web technologies - development velocity would definitely be higher there). But on the flipside, given all of the available LLM and AI support that we are privileged to have today, I figured the velocity should be comparable depending on how much I leaned on those. And lean I did. Godot/Rust seemed the better call to me and my intent anyway - going with the 'it's not just the end but the journey that matters' philosophy. So yes, there has been heavy use of LLMs & AI to generate a lot of the code. But I do review and steer actively, not relying solely on vibes, and there's a few bits here & there that have been human authored. There are definitely a lot of polish and QoL items that need to land to make the end user experience better, but in the meantime, let me know your thoughts and/or concerns! Repository: https://ift.tt/TtNO4wQ Docs/Blog: https://godot-pty.github.io/gpty/ https://ift.tt/TtNO4wQ September 11, 2026 at 09:33PM

Show HN: Raiders at the Gate – tower defense web game https://ift.tt/jRcnah8

Show HN: Raiders at the Gate – tower defense web game https://ift.tt/JCgv1ql September 10, 2026 at 11:26PM

On the Right Track with the T Third

On the Right Track with the T Third
By

Thousands of riders depend on the T Third to get around San Francisco every day. Your ride on one of San Francisco's most important transit lines has gotten faster and more reliable. The T Third crosses the city to connect Bayview-Hunters Point and Visitacion Valley to places like Chinatown, Union Square, Mission Bay, SoMa and Dogpatch. And the T Third Improvements Project has reduced delays on the line. Providing you with a better trip Long travel times have been a problem on the street-level section of the T Third. And that can make your trip less reliable. The T Third Improvements Project...

Published 2026-09-09T00:00:00Z
https://ift.tt/tJjsYOr

Show HN: I'm not a PostgREST fan this is why I'm building an alternative https://ift.tt/8WR90iL

Show HN: I'm not a PostgREST fan this is why I'm building an alternative Don't get me wrong, core PostgREST concept is good. But I do think that stuff like Haskell and PlSQL should not be used in production. This is why I came up with the idea to make my own API generation service. Right now it is using SqLite as a main DB and generates endpoints based based on TOML DSL with some LUA scripting. I guess in several week I will have fully working usable MVP. https://ift.tt/IbGPUgl September 9, 2026 at 11:17PM

Show HN: Pipnote – Share a reminder they can't peek at until it arrives https://ift.tt/J4DjRGH

Show HN: Pipnote – Share a reminder they can't peek at until it arrives https://pipnote.app/ September 9, 2026 at 01:18AM

Show HN: Linubot – like Grok bot and Hermes had a kid https://ift.tt/Zqrz3FM

Show HN: Linubot – like Grok bot and Hermes had a kid Work on Linux and Android with Tailscale. Because we also deserve some good stuff before everyone else. Any provider, or providers. Desktop workspace controlled with IPC by the agent, memories, skills, MCPS, self-learning, and more. Friendly UI and advanced area for power users. https://ift.tt/T7MHWrE September 9, 2026 at 12:14AM

Show HN: Gote – a CLI note-taking management tool for plain Markdown https://ift.tt/mQTWXxK

Show HN: Gote – a CLI note-taking management tool for plain Markdown Repo: https://ift.tt/F9H1gM0 I built a note taking management tool to use at work for quick note taking and searching. It is intended to be quite minimal. Another focus was to have lots of shortcuts and to keep things close to the homerow for ergonomics and speed. Notes are just markdown files in whatever directory you point it at. The only thing it keeps elsewhere is an index in ~/.gote. If you stop using it you still have all your notes. Tags go on the first line of the note, separated by periods, like .project.urgent.work. No frontmatter. Search is built into the binary, BM25 with stemming, so there's no ripgrep or fzf to set up. gote s meeting searches titles and content, gote s -t .work filters by tag, gote s -w 2412 pulls up notes from December 2024. Most commands have a two letter version that combines a list with an action. ro is recent + open, sv is search + view, tp is tag + pin. When results come up you pick one with a home row key instead of arrows or numbers. Binaries for mac/linux/windows are on the releases page, or go install github.com/banorton/gote@latest if you have Go. The goal of this project was not to learn something new or build something impressive. I simply had a need and filled it with a vibe-coded solution. I have been using it for about a year now and it does more or less exactly what I wanted. So I thought I'd share in case others are interested in using it, contributing to it, or want to use it as a reference for their own solution. September 7, 2026 at 10:25PM

Show HN: Possess, a TUI to browse and transfer coding sessions https://ift.tt/gClXwkE

Show HN: Possess, a TUI to browse and transfer coding sessions https://ift.tt/Y5fHR2d September 7, 2026 at 11:43PM

Show HN: Sol, my macOS music player and jukebox app, is now free and open source https://ift.tt/nDT4Q6L

Show HN: Sol, my macOS music player and jukebox app, is now free and open source https://ift.tt/i2NnOTK September 7, 2026 at 01:54AM

Show HN: PixelDraw – simple drawing software for web(WASM) and desktop https://ift.tt/7UYNMGx

Show HN: PixelDraw – simple drawing software for web(WASM) and desktop https://ift.tt/1ptAkNo September 6, 2026 at 09:23PM

Show HN: Cowbull – Wordle-like game where you only get match counts https://ift.tt/lbhWvsN

Show HN: Cowbull – Wordle-like game where you only get match counts CowBull is a word game where you have to find the hidden four-letter word. You make guesses and get back the count of exact letter matches (bulls) and partial matches (cows). It’s a bit like Wordle, but you are not told the exact letters which match, only the count of matching letters. Scoring is based on the number of guesses and time used. There is a single player mode and a multi-player mode (no account required). The game source code is at https://ift.tt/c2nysQH . When I was younger, we used to play this game on paper at the back of the classroom during school and college in India. During Covid, I made this online version of the game to play with my kids. https://cowbull.co/ September 6, 2026 at 06:07AM

Show HN: I made Blocknado, an orbital stacker. Think Tetris with square rings https://ift.tt/9keEQKc

Show HN: I made Blocknado, an orbital stacker. Think Tetris with square rings I've loved stacker type games for as long as I can remember. But I've always wanted a version with a bit more strategy and puzzle dynamics. Blocknado is a game I've always wanted to play, maybe others will too. https://ift.tt/qBnJXxE September 6, 2026 at 05:46AM

Show HN: Merge Wikipedia articles across languages into one page https://ift.tt/ERUGfNn

Show HN: Merge Wikipedia articles across languages into one page https://ift.tt/6kers85 September 6, 2026 at 02:18AM

Show HN: Fast Cut Video tool for cutting video for Agents https://ift.tt/8QNLeWK

Show HN: Fast Cut Video tool for cutting video for Agents Hi HN, Build this tool in a couple of hours with OpenAI Codex to scratch an itch. As part of my videos workflows I needed to cut the videos myself since the agent is pretty bad cutting and timing using only the transcription. But using a big app like Premiere or DaVinci Resolve was overkill so I used Astra to build an app tailored to that very task. It will be useful to others with the same problem. Tested only on Silicon Macs, so any feedback is welcome. As part of my workflow when the agent get the video it can open the app with the video or videos loaded in the timeline for me to cut and then export the cuts in a format the agent can use and continue the workflow. https://ift.tt/ma91Cfx September 6, 2026 at 02:36AM

Show HN: Coder Eval – A Framework for Evals https://ift.tt/qB9DYMJ

Show HN: Coder Eval – A Framework for Evals We needed a framework to write our evals in and easily update them, run A/B tests, set all kinds of constraints (ex: timeouts, number of turns), and configure the execution environment (ex: sandboxes, dependencies, how to handle AskQuestion). We also have an agent judge. It’s all in a YAML file now. We seem to be moving toward a world where companies rely more and more on evals to decide what to ship, so this should be super helpful for normalizing evals across teams. https://ift.tt/R6F5bgI September 5, 2026 at 12:39AM

Show HN: TERMy – A fast terminal assistant that does not use LLMs https://ift.tt/IKuSDxg

Show HN: TERMy – A fast terminal assistant that does not use LLMs I love research and development, you may have heard of me because of PJON (Padded Jittering Operative Network). It is a network protocol I started developing in 2010, which was recently implemented in silicon by the ETH Zurich university thanks to the research of Pius Sieber. I am excited to share with you TERMy, a terminal assistant built on top of the NPC-Forge framework. Unlike everything else being built today, TERMy does not use embeddings, machine-learning or LLMs. It runs on the CPU (even on a Raspberry Pi Zero) both in the terminal or client-side in a browser tab and responds in milliseconds. It is a cynical but very knowledgeable Linux terminal assistant that translates your natural language into shell commands without relying on a single artificial neuron. I had a chance to focus for 2 months on my personal projects since early July, during the strange times of AI price hikes and the end of subsidized tokenmaxing. I was curious to see if I could develop from scratch a terminal assistant capable of handling simple natural language requests. I have a bad memory and got used to ask to copilot "activate the virtual environment" or similar trivial operations spending a non negligible sum every month. I started thinking, maybe I can do something to make my workflow more efficient? Do I really need trillions of parameters to accomplish those tasks? How it Works When you type a prompt, it goes through a lightweight NLU pipeline written in ~1000 lines of Python that implement the following steps: 1. Strip expletives, interjections, encouraging, discouraging and thanking words (remove noise) 2. Sentiment analysis 3. Exact Match (very fast) 4. Template Match (slower) 5. Probabilistic Match (even slower) Step 5 relies on: 1. IDF (Inverse Document Frequency) to identify rare words. 2. BOW (Bag Of Words) to accommodate word inversions. 3. IDF weighted Levenshtein to safely handle typos. Permission gating is hardcoded into the dataset and enforced for all potentially destructive commands, so it's inherently safer than letting an unpredictable LLM run wild on your machine. - TERMy in operation: https://www.youtube.com/watch?v=qeIp0xePLBg - Variance and typo tolerance: https://www.youtube.com/watch?v=tQvGDk6fkk0 - Copilot integration: https://www.youtube.com/watch?v=Wzzouhq2a8A - Advanced features: https://www.youtube.com/watch?v=qeIp0xePLBg - Source Code: https://ift.tt/HKc7avw https://ift.tt/IKOyoTD September 4, 2026 at 02:33PM

Show HN: Sirenfall, a civil defense siren synthesized in Web Audio https://ift.tt/30tOjT9

Show HN: Sirenfall, a civil defense siren synthesized in Web Audio I built a cool browser scene, a shelter window over a dystopian city in the rain. Pull the breaker and a siren tower spins up. The wind-down is super unsettling and I love it. The siren is not a recording. Two port rings make a 5:6 dual tone, 515 and 618 Hz. The horn turns at 4 RPM and everything (loudness, brightness, doppler and stereo position) follows the horn angle you can see on the tower. Rain and thunder are the only recordings. Build it yourself, source is one HTML file and MIT: https://ift.tt/MhteJyV How it works: https://sirenfall.live/how-it-works Press F for fullscreen, S to save a picture. Landscape on phones. No cookies, no personal data, the only request besides audio is an anonymous page-count pixel. Appreciate any feedback and shoutout to the siren community! https://sirenfall.live/ September 4, 2026 at 02:14AM

Show HN: NGPDFs – 50 PDF tools that never upload your files https://ift.tt/286Rdg1

Show HN: NGPDFs – 50 PDF tools that never upload your files https://ngpdfs.com September 3, 2026 at 11:13PM

Show HN: Arbitale – emergent choose your own adventure game https://ift.tt/Y3R5KGX

Show HN: Arbitale – emergent choose your own adventure game Hi everyone! I am building Arbitale and it is currently in pre-alpha. I'm looking for play-testers who can give feedback on the product. The goal is to eventually have different characters in the game be their own agents, with emergent narratives that are totally unique to each playthrough. Each agent acts according to its own pre-written incentives. Would love to hear what you think! https://ift.tt/8FjTDQY September 3, 2026 at 12:12AM

Show HN: SufferTrails, the Anti-AllTrails https://ift.tt/ArtsGgO

Show HN: SufferTrails, the Anti-AllTrails Don't get me wrong. I love a good family-friendly hike as much as anyone else. But sometimes you want to be grimacing, not smiling, on the trail. SufferTrails is that, and not much else. All trails not allowed. https://ift.tt/Uyhe2Em September 2, 2026 at 11:27PM

Show HN: AQ – A multiplayer workspace for coding agents on your own cloud https://ift.tt/mbFW5tC

Show HN: AQ – A multiplayer workspace for coding agents on your own cloud https://aq.dev/ September 2, 2026 at 11:16PM

Show HN: Newton's Orchard – Browser-based space/gravity playground https://ift.tt/mlkdS1M

Show HN: Newton's Orchard – Browser-based space/gravity playground Hi HN! My son spent a lot of time in 5th grade playing with the 2-dimensional PhET gravity/space sims[1]. His STEM class uses it, and he was playing with it at home, and has been looking for more. I found several space sims/playgrounds online, including NASA's Eyes on the Solar System[2], but nothing that seemed to offer a real progression from what he was getting out of the PhET sims, so I built Newton's Orchard over the summer for him. https://ift.tt/g0LIwiT The source is also published at https://ift.tt/SiPt07a It features deterministic time scrubbing, all bodies are editable, different preset systems and experiments, missions to complete (complete 3 to unlock the blackhole object type and mission), and a gallery for user submissions. I would very much appreciate feedback from HN - particularly from the lens of making it engaging and illuminating for students, and anyone that would like to just get a feel for how gravity works. Thank you! [1] https://ift.tt/WSpqDP2... [2] https://ift.tt/I4lrZqR https://ift.tt/g0LIwiT September 1, 2026 at 10:13PM

Show HN: Mcptunnels – ngrok for MCP with basic OAuth https://ift.tt/JC37OTS

Show HN: Mcptunnels – ngrok for MCP with basic OAuth Basically, OpenAI tunneling sucks and is broken most of the time. This can get you tunneling pretty trivially to share with someone temporarily. Roadmap includes being able to host MCP servers trivially as well. https://terragohan.github.io/mcptunnels/ September 2, 2026 at 02:04AM

Show HN: Semantic Overlays – an NX bit for LLM prompt injection (live demo) https://ift.tt/t6O27pu

Show HN: Semantic Overlays – an NX bit for LLM prompt injection (live demo) I've built a new method for steering LLMs called Semantic Overlays, small trained adapters on a frozen model which change how it perceives a piece of its context. The most readily applicable usage is to mitigate prompt injection, and it lets us take a very-injectable Qwen-3.5-9B to SOTA scores on all the prompt injection benchmarks I could find. (They are only blackbox attacks, but I did NOT train on anything like them — whitebox attacks are out of scope for this paper) I'm excited for you to play with the tech — see if YOU can break it! (let me know if you can) Paper at https://ift.tt/BAFRDpZ if you want to read more about it, code at https://ift.tt/EWKauSN , adapters at https://ift.tt/3zNdyMC... Also https://ift.tt/jKtpiJI if you wanna watch a little video I made! https://semantic-overlays.vercel.app/ September 1, 2026 at 11:10PM

Show HN: Guess the AI model from its web design https://ift.tt/cQp81Lx

Show HN: Guess the AI model from its web design https://ift.tt/rSuHnR3 September 1, 2026 at 01:10AM

An Operator's Perspective on How Muni Moves and Connects People

An Operator's Perspective on How Muni Moves and Connects People
By

Sincere Dow steps out of the operator chair to strike a pose on a Muni train. For Muni Operator Sincere Dow, no two days on the job are quite the same. “I pretty much float and drive all the train lines,” she said. “But mainly I am on the F Market. And I also do the J Church a lot.” Dow’s work takes her through unique neighborhoods and by some of San Francisco’s most iconic destinations. There also are schools on some of her lines. And she loves helping students get where they need to go. “They’re all very grateful,” Dow said. “They’re especially appreciative when you take that extra time to...

Published 2026-08-31T00:00:00Z
https://ift.tt/9gW8tSJ

Show HN: Corporate Mind Games – logic puzzles with a sarcastic corporate theme https://ift.tt/MVtLsgI

Show HN: Corporate Mind Games – logic puzzles with a sarcastic corporate theme I have always enjoyed building and playing puzzles. I am a regular user of NYT Games and LinkedIn Games. I thought it would be fun to create some corporate-themed puzzles as a silly alternative to LinkedIn Games. So, I built Corporate Mind Games. The games are intended to be a little bit sassy and sarcastic, but I hope they are enjoyable and challenging. I was planning to post this on HN last week. However, many years ago, I made a game called Don't Wordle, which unexpectedly appeared on the front page last week. I ended up getting distracted by that, and I also figured it was more appropriate to post this game on HN on a Monday morning given the theme. So here we are. https://ift.tt/mW245wC August 31, 2026 at 09:38PM