github.com
Show HN: MAItion – Open-source RAG with pluggable connectors and chat UIhttps://github.com/WikiTeq/mAItionHey HN,We wanted to share a new tool we’ve been working on.Even when documentation is well-structured, sometimes it’s hard to find what you need. Information ends up spread across wikis, cloud drives, chat tools, ticketing systems, etc., and search can break down as info grows.mAItion is our attempt to solve that. It’s a way to unify search and retrieval across the systems that teams already use. It's an open-source alternative to tools like Glean and GetGuru, but it’s designed to be flexible, lightweight, self-hostable, and easier to adapt.What it does:- Ingests data from all the connected sources and builds vector and BM25 index- Converts the data into markdown (using MarkItDown, configurable)- Lets you run Hybrid search across the data stored via API or MCP- Lets you chat with the data using Open Web UI as a frontend (bundled) or connect other arbitrary chat UI- Keeps the data fresh using scalable Celery workers (per source schedules)- Deduplicates documents over sources during ingeApr 29, 2026 10:26 PM
bing.com
Kool-Aid Ventures Into New Territory With First-of-Its-Kind Drink Mixhttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8ba715dd7f47a99b323b39e1c94024&url=https%3A%2F%2Fwww.yahoo.com%2Flifestyle%2Farticles%2Fkool-aid-ventures-territory-first-155942720.html&c=16735642321991396925&mkt=en-usKool-Aid has long been synonymous with sweet drink mixes, echoing memories of our youth as the drink of choice on summer days spent running through sprinklers. But now, the brand is venturing into ...Apr 29, 2026 8:59 AM
world.gjlmotea.com
Show HN: Mvndvs TERRARUM – A Renaissance-style 3D world maphttps://world.gjlmotea.com/In this fast-paced, AI-exploding era, I felt a strong desire for a slower, more deliberate pace.I built a 3D interactive world map: https://world.gjlmotea.com/Lately, time seems to be flying by. I've been building a lot of quirky, experimental projects, but honestly, my mood has been quite restless and low. There are just too many options and new technologies emerging every day—choice paralysis sets in before I can even figure out what I truly want to build.But this project is different. It’s more grounded, human, and carries a sense of warmth. I wanted to create an immersive experience with a Renaissance, antique parchment aesthetic.Features:Explore: Browse different countries and click on nations or architecture for detailed information.Data Layers: Toggle between various global map modes including Architecture, Culture, Academia, Commerce, and Military.Themes: Multiple visual themes to switch between.3D Architecture: Scale the models of buildings (like Taipei 101) to your liking.GloApr 28, 2026 4:13 PM
news.ycombinator.com
What type of code should you generate with AI?https://news.ycombinator.com/item?id=47917192I (and many people) have been thinking a lot about "what are the best tasks to do using AI." My personal framework I've been using is, you should use AI to generate code when: 1. The code is easy to validate 2. It is not important that a human understands itI find that the best thing to AI generate is something on the level of a pure function (easy to validate + not necessarily important to understand implementation as long as you understand the interface). I've tried doing things like generating whole services or applications, but those often violate rule 1 - it's hard to validate an entire application behaves "correctly" when "correctly" isn't really well defined - i.e., are there memory leaks, is it secure, is it able to be monitored, etc.I'm curious of others thoughts on this topic. The larger the task you complete using AI, the more time savings, but also the more likely that a catastrophic error exists in the generated code. A side effect of following these rules is there is not Apr 27, 2026 2:50 AM
github.com
Show HN: A Karpathy-style LLM wiki your agents maintain (Markdown and Git)https://github.com/nex-crm/wuphfI shipped a wiki layer for AI agents that uses markdown + git as the source of truth, with a bleve (BM25) + SQLite index on top. No vector or graph db yet.It runs locally in ~/.wuphf/wiki/ and you can git clone it out if you want to take your knowledge with you.The shape is the one Karpathy has been circling for a while: an LLM-native knowledge substrate that agents both read from and write into, so context compounds across sessions rather than getting re-pasted every morning. Most implementations of that idea land on Postgres, pgvector, Neo4j, Kafka, and a dashboard.I wanted to go back to the basics and see how far markdown + git could go before I added anything heavier.What it does: -> Each agent gets a private notebook at agents/{slug}/notebook/.md, plus access to a shared team wiki at team/.-> Draft-to-wiki promotion flow. Notebook entries are reviewed (agent or human) and promoted to the canonical wiki with a back-link. A small state machine drives expiry and auto-archive.-> Per-eApr 25, 2026 8:53 AM
notationchesscoach.app
Show HN: Notation – an iOS chess coach (Stockfish on device and optional BYOK)https://notationchesscoach.appHi HN. Solo iOS middle aged, slightly burnt out dad from the UK here, wondering if anyone is interested in what I've been building. Lifelong chess fan, not very good but always trying to improve and I started to build Notation: Chess coach a while back. It's been live a month now on the App Store, doing modest numbers, and I've been learning a lot... mainly make my add brain slow down, stop, test everything twice and never when tired. Submissions will always wait another day!Many people play on chess com, or lichess, just as many in apps and OTB. What my app does (differently I hope) is bring that data together.You can import PGNs of your games, connect to chess com and lichess and sync your games over or play in the app. If you have a Chessnut electronic board, you can play IRL on that as well.Every game, every move is then analysed by a series of on device detectors, and classifiers. And I'm not talking about"I run Stockfish 18 and get your move evals" like most tools do.... Theres aApr 22, 2026 7:46 AM
charlielabs.ai
Show HN: Daemons – we pivoted from building agents to cleaning up after themhttps://charlielabs.ai/For almost two years, we've been developing Charlie, a coding agent that is autonomous, cloud-based, and focused primarily on TypeScript development. During that time, the explosion in growth and development of LLMs and agents has surpassed even our initially very bullish prognosis. When we started Charlie, we were one of the only teams we knew fully relying on agents to build all of our code. We all know how that has gone — the world has caught up, but working with agents hasn't been all kittens and rainbows, especially for fast moving teams.The one thing we've noticed over the last 3 months is that the more you use agents, the more work they create. Dozens of pull requests means older code gets out of date quickly. Documentation drifts. Dependencies become stale. Developers are so focused on pushing out new code that this crucial work falls through the cracks. That's why we pivoted away from agents and invented what we think is the necessary next step for AI powered software developmApr 21, 2026 4:16 PM
news.ycombinator.com
Ask HN: Would you use revocable digital signatures to verify AI/Other content?https://news.ycombinator.com/item?id=47848539I’ve been exploring a potential product direction and wanted to sanity check it with people who actually build and ship things.Background: I’ve been working on a system using our core tech that can generate and verify digital signatures, but with a slightly different property than traditional approaches. The signatures are natively revocable. If the underlying model/system shouldn’t be trusted anymore, the signatures can be revoked either through a hard (delete the signing model) or soft (revoke the lease for the signing model) mechanism. I believe this feature is very beneficial on its own, but there are several other interesting properties (no key management, distributed verification, embedded metadata, etc.)Originally this came out of some deeper R&D work we’re doing, but I’ve been thinking this might actually be the most practical “wedge” into the market while we continue that research (and fund the research).One area that’s been interesting is applying this to AI systems. SpecificApr 21, 2026 1:28 PM
github.com
Show HN: Mulder – Containerized MCP server for digital forensics investigationshttps://github.com/calebevans/mulderI built Mulder for the SANS "FIND EVIL!" hackathon. My background is not in forensics, but I was interested in building the agent. I've been seeing some really cool results and thought this community might like it!Mulder is an MCP server that exposes forensic tools like Volatility, Sleuthkit, Plaso, and more as typed tool calls. Every invocation is recorded in an append-only audit log. When an agent finds something, it submits a "finding" to the case database. Each finding must have evidence citations that are validated before being admitted. If the agent cites evidence or a tool execution that doesn't exist, the finding is dropped.The biggest problem I needed to solve was context window pressure. These investigations often contain a lot of data, and processing it produces a lot of noise that will fill a context window quickly. To avoid this, all tool output is stored in a per-case SQLite database with full-text search. The agent can then search the database and request segments of tooApr 21, 2026 1:06 PM
news.ycombinator.com
Getting back into photography, ditching the phone camera in 2026?https://news.ycombinator.com/item?id=47842292I've been looking through twenty-five-odd years of my own photos. The collection includes scanned 35mm and medium format images; digital pictures from a few Canon Powershot generations and a 20D SLR; and about five iPhones.I've noticed that the non-cell-phone pictures tended to be better, and that in general I seemed to have quite a bit more fun with photos taken with, well, "real" cameras. Probably the best ones were taken with the 20D, which for its time was a really nice camera, for which I was able to bring over some lenses from my film days.I wouldn't rule out more film photography but I think of that as a somewhat separate track. I'd like to get a digital camera that captures a bit more of what I enjoyed about film photography - that high image density, looking through a physical viewfinder, not necessarily curating images in real time... looking, shooting, and moving on.I've been following Fujifilm cameras for awhile in this market, having read about them here a few years ago. TApr 20, 2026 11:10 PM
bing.com
Attendee reveals what really went down at Meghan’s Sydney retreat — and how Harry stole the showhttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a898042179349db91ee85fed99de0f6&url=https%3A%2F%2Fwww.msn.com%2Fen-us%2Fentertainment%2Fcelebrities%2Fattendee-reveals-what-really-went-down-at-meghan-s-sydney-retreat-and-how-harry-stole-the-show%2Far-AA21m5XL&c=10741498966449288107&mkt=en-usMeghan Markle's ‘Her Best Life’ retreat in Sydney had been in the news since it was announced—and not always for the right reasons. The event drew scrutiny before it even began, with critics ...Apr 20, 2026 9:38 PM
github.com
Show HN: Sfsym – Export Apple SF Symbols as Vector SVG/PDF/PNGhttps://github.com/yapstudios/sfsymI found myself reaching for SF Symbols' 'Copy Image As…' quite often during agentic design sessions, so I made a command-line tool that the agent can use by itself. It exports Apple SF Symbols as SVG, PDF, or PNG.The vector paths come directly from macOS's symbol renderer. Internally it reaches a private ivar on NSSymbolImageRep to get the CUINamedVectorGlyph, draws into a CGPDFContext, then walks the PDF content stream back out as SVG `d` commands. The output matches what the system draws, rather than an approximation traced from rasters.A few things about it:- Every subcommand accepts `--json`, and `sfsym schema` returns a machine-readable description of the whole CLI. - Symbol enumeration reads the OS's Assets.car BOM tree, so the list of 8,300+ names stays current with macOS updates without a version table in the binary. - Each SVG `` carries a `data-layer` attribute, so you can retheme in CSS without touching geometry.It's been saving me a bunch of clicking. Please let me know if Apr 18, 2026 3:44 AM
bing.com
Xbox is Giving Some Lucky Gamers Free Gift Cardshttp://www.bing.com/news/apiclick.aspx?ref=FexRss&aid=&tid=6a8b221ff4bd4d219b7361f7586ffdb0&url=https%3A%2F%2Fwww.military.com%2Foff-duty%2Fgames%2Fxbox-giving-some-lucky-gamers-free-gift-cards.html&c=14981452640059088269&mkt=en-usXbox gamers may be able to get some free or discounted games courtesy of a new freebie granting them gift cards to the Xbox Store. Lucky Xbox gamers are discovering that they've been given a free gift ...Apr 17, 2026 1:36 AM
spacenews.com
Artemis 2 is a benchmark in our space exploration program — and it depends on steady NASA fundinghttps://spacenews.com/artemis-2-is-a-benchmark-in-our-space-exploration-program-and-it-depends-on-steady-nasa-funding/Now that the Artemis 2 mission has been successfully completed, it’s worth taking a look at where NASA stands on the role of humans in exploring space and what its path forward should be. Doing this is especially important today, with some people questioning on line whether we can afford the cost of sending humans […] The post Artemis 2 is a benchmark in our space exploration program — and it depends on steady NASA funding appeared first on SpaceNews.Apr 16, 2026 1:00 PM
news.ycombinator.com
Ask HN: What do founders want from a one person dev studio?https://news.ycombinator.com/item?id=47782421I’m a full stack developer and recently started a small solo studio:https://www.launchcraft.studio/The idea came from a pattern I kept seeing with early-stage founders:They don’t usually struggle with ideas.They struggle with execution speed.Either they:-hire agencies that take months-or try to build themselves and get stuck midwayI’ve been focusing on helping with:-MVPs-landing pages-small SaaS productsMostly handling everything end-to-end (frontend, backend, deployment).What I’m still trying to figure out is:What do founders actually expect from a setup like this?Some tradeoffs I’ve noticed:-speed vs long-term scalability-polish vs shipping quickly-fixed scope vs iterative buildsIf you’ve worked with freelancers / agencies / solo devs before:-what worked well?-what didn’t?-what would make this actually valuable to you?Trying to refine this into something genuinely useful rather than just another “dev service”.Apr 15, 2026 5:35 PM
news.ycombinator.com
Why Vibe Coding Failshttps://news.ycombinator.com/item?id=47778946i am using claude to maintain an agent loop, which will pause to ask for users' approval before important tool call. while doing some bug fixes，i have identified some clear patterns and reasons why vibe coding can fail for people who dont have technical knowledge and architecture expertise.let me describe my workflow first - this has been my workflow across hundreds of successful sessions: 1. identify bugs through dogfooding 2. ask claude code to investigate the codebase for three potential root causes. 3. paste the root causes and proposed fixes to claude project where i store all architecture doc and design decision for it to evaluate 4. discuss with claude in project to write detailed task spec - the task spec will have a specified format with all sorts of test 5. give it back to claude code to implement the fixin today's session, the root cause analysis was still great, but the proposed fixes are so bad that i really think that's how most of vibe coded project lost maintainability Apr 15, 2026 1:50 PM
jimmont.com
Show HN: HWT (Hash Web Tokens) – minimalist protocol for auth statehttps://www.jimmont.com/hwt/My frustration solving auth with JWTs led to reinventing the wheel for more predictable token integrity and transparency, while improving flexibility. The mix of features separates concerns more cleanly than what's been available, allowing higher throughput, custom codecs, delegation to both services and domains while easing key rotation and other practicalities.The design is focused on the token as state guarantee, not creation, revocation and the range of other separate concerns and application responsibilities. The spec conventions attempts to ease development with jurisdiction/data sovereignty and authorization in the authz field. It also enables and eases delegation between services and domains without centralized service exposure. The implementation is standalone and has demos for Deno, Cloudflare, etc in https://github.com/hwt-protocolFeedback and critique of the security logic and approach appreciated.Apr 15, 2026 1:24 PM
nasa.gov
NASA Receives 7 Nominations for the 30th Annual Webby Awardshttps://www.nasa.gov/general/nasa-receives-7-nominations-for-the-30th-annual-webby-awards/Since it began in 1958, NASA has been charged by law with spreading the word about its work to the widest extent practicable. From typewritten press releases to analog photos and film, the agency has effectively moved into social media and other online communications. NASA’s broad reach across digital platforms has been recognized by the […]Apr 14, 2026 8:26 PM
news.ycombinator.com
Show HN: BNNR – a closed-loop pipeline for improving vision modelshttps://news.ycombinator.com/item?id=47719542Hi HN,we’ve been working on computer vision models for a while, and one thing kept coming up: improving them is surprisingly unstructured.You train a model, try a few augmentations, tweak some hyperparameters, run it again - and if the metric goes up, you keep it. But it’s often unclear why it improved, or whether the model is actually learning something better.We kept running into questions like:- Did this change actually help, or is it just noise?- Is the model focusing on the right features?- Are we improving generalization, or just overfitting differently?So we built BNNR (Bulletproof Neural Network Recipe) - an open-source PyTorch toolkit that turns model improvement into a closed loop:- Train a controlled baseline- Explain what the model actually learned (via XAI methods like OptiCAM / GradCAM)- Improve by testing candidate strategies in parallel- Prove the result with structured comparisonsOne thing we focused on is avoiding “blind” changes. Instead of committing to a single ideApr 10, 2026 3:24 PM
lingle.ai
Show HN: Lingle – Voice agent to simulate zoom-based personal language lessonshttps://lingle.ai/In my opinion the best way to learn a language (outside of moving to a different country) is to get a personal tutor and have consistent 1 on 1 lessons. I used Preply or iTalki for this back in the day but had issues with flexibility and pricing.I've been trying to simulate the experiences that I had on those platforms with a voice agent.You can try a demo for free here (or watch a video of me using it on the landing page):https://lingle.ai/tryout/lingle-showcaseRight now the agent can plan a lesson with visuals and a conversation focus, guides the user through the lesson - correcting and explaining things on a whiteboard if necessary, and asks relatively engaging questions. There is also a long term user model that builds over time, that maintains memory of the user's produced vocabulary, grammar, and skills, meant to make each lesson informed by the past and build on the user's knowledge (obviously this is still relatively fragile in practice).The voice agent itself has actually deceApr 9, 2026 10:28 PM
