Show HN: Supapool – a Supabase per coding agent in ~400 ms hi HN, I built supapool.io, an ephemeral full copy of supabase's services that you can spin up in ~400 ms (Auth, postgres, storage, realtime). so if you run multiple coding agents in parallel in different worktrees, they can now have their own copy of supabase without making changes that conflict with eachother. > why not use supabase docker locally? when I run 3-4 instances locally, my macbook gets hot and sometimes freezes. > why not use supabase branches? branches take minutes to setup, and are designed for persistence. this is expensive, and for a dev environment, it is too slow. > why not use mocks? mocks are bad for agents. i expect agents to test their migrations, SQL against real prod service behavior. agents hallucinate working mocks often. However, upside of mocks is that its faster and runs locally, but with supapool, the upside is less convincing. > how does it work/how is this economically viable? starting supabase in 400ms requires a few things:
1. a pool of ready supabase instances running warm, and colocated with region failover (us-east, us-west, europe-west, asia-southeast)
2. fast autoscaling when pool starts to shrink with microVM/firecracker
3. gutting strong persistence guarantees. dev agents don't need WAL, fsync, PITR, replication. anything for HA on a ephemeral supabase instance is bloat its in beta right now, and i'm using our gcp credits to bankroll this, so its free. the eventual pricing will be something like $/instance second and more cost effective than branching or self hosting/maintaining a supabase cluster. would love to get your feedback if you use supabase, and if you think there's something better that would fit your local coding agent setup. Thanks! https://supapool.io/ July 29, 2026 at 11:04PM
Show HN: Hacker Fables – A satirical cyberpunk novel you can read as a man page The project is entirely open source. It takes a single Pandoc Markdown source file and produces HTML, EPUB, PDF, man pages, GNU Info, an audiobook, an HTML-integrated audio player that highlights the content as it plays, and an MP4 output for the "video audiobook." It's generic enough to be reusable. I plan to use the pipeline for my tech blog later on. The novella contains a lot of programming lore. It's an over-the-top love letter to the old-school Linux programmer and to the craft itself. Source code: https://ift.tt/4lkVxu3 https://sebastiancarlos.github.io/hacker-fables/ July 30, 2026 at 07:22PM
Show HN: CheapFoodMap – A map of good meals under $10 I was recently laid off after 18 years, and gave myself 100 days to build soething useful in public. CheapFoodMap is a crowdsourced map of meal under $10, excluding franchises, local good eats only. It's inspried by 거지맵 (Begger's Map) a Korean crowdsourced map students use to find cheap eats. Ocverage is heaviest in Texas, since I live in Dallas, but have 1200 meals across 15 US cities. Seed data came from Google Review, 4.2 star or higher with at least 500 reviews, and verified price under $10 per menu item. Things I would love feedback on : whether the price-freshness model makes sense, and what would make you trust the price on a site like this. How to encourage people to update prices, since inflation is making food price very frequent. https://ift.tt/HG86loq Any and all suggestion will be super helpful. Thank you! https://ift.tt/u4POnRy July 29, 2026 at 10:29PM
Show HN: Writekin – fine-tune a local LLM on your own writing, on your Mac Hey Hacker News! I built Writekin over the past week because I was tired of AI writing that didn't
sound like me, even though I had just used AI to clean it up, rather than wholesale write it. The usual fixes I found online for this were: - Some sort of SKILL.md, or - A system prompt full of rules to strip the generic AI tells (e.g. no em-dashes, none of the stock phrases, varying the sentence length, etc). While those cleaned up the surface a bit, Pangram still came back as ~100% AI written, which was frustrating, as again it was mainly taking my sloppy copy and tweaking it. So when building Writekin I took a different route: Writekin fine-tunes a local model on your own writing. It reads what you've already written (Apple Mail, iMessage,
local documents, chat exports), curates it into a training corpus, and
runs a QLoRA fine-tuning on-device via Apple's MLX. A Compose screen then
drafts and rewrites in that voice. Everything runs on your Mac. Ingestion, training, and generation are all
local. The only network calls are: (1) When you download the model weights from Hugging Face and (2) The Sparkle update check. Training on your own Mail/Messages only felt okay to ship if the end user could verify that, so the source is public — so you can read exactly what it does! Quick gut check: It's v0.9 and the output is uneven. Honestly, sometimes it nails your
voice, and sometimes it's just completely off. This is more a "this is possible and
kind of works" than a finished product. Would genuinely love feedback! Source: https://ift.tt/hNE0Dqx https://ift.tt/hNE0Dqx July 29, 2026 at 12:35AM
Show HN: Somebodyhire.me I originally had this as a personal resume site but decided to build it out into a platform where anyone can sign up and host a resume page.
Once I build up the talent pool, I plan on marketing to hr depts and hiring managers. I just don't know a single person who's happy with hiring on either side right now and this feels like a better solution to me. https://ift.tt/HxTg8Dv July 28, 2026 at 09:48PM
Show HN: Let's Seal – Let's Encrypt for document signing, free and self-hosted TLDR, Let's Seal gives the finger to Adobe and every doc signing tool (docusign, google, etc) who pay to play with the Adobe Approved Trust List and then charge you for something that should be free. Currently even the person checking if a document/contract is sealed or code is authentic has to also be inside the same Adobe walled garden too. Verification, the part that should be free is the part everyone charges for. Thats the shape Let's Encrypt fixed for TLS, and I wanted the same thing for documents and files. The core idea therefore needed to go a bit beyond e signatures and i created an open standard (SEAL), plus free tools that implement it. When you seal a file, three independent things happen. 1. it gets a signature from a certificate authority, chaining to a public root.
2. its record is appended to an RFC 6962 transparency log. and
3. its SHA256 is timestamped on a public blockchain (Bitcoin) via OpenTimestamps.
Those three give you integrity, transparency and a timestamped proof. And importantly, none of those depend on Let's Seal and none are gated. You can verify with the tools you already have, no Let's Seal account and no Let's Seal software. A sealed PDF carries a standard PAdES signature, so any PDF reader validates it. A sealed build artefact carries a cosign compatible signature and a SLSA provenance attestation. The Bitcoin timestamp verifies with stock ots. 3 ways to use it. 1. The free web app. We kindly have backing from Backblaze to cover storage costs for the foreseeable. So you can upload or issue any number of documents, get a public proof page at /d/ and verify it at https://ift.tt/D6CjL7E for free. Multiple accounts, multiple seats, enterprise functions. Free. 2. Self host the whole thing. Apache-2.0, one Next.js app plus a signing service that holds the CA key on localhost. Storage is any S3-compatible bucket or local disk. If you'd rather run your own root of trust, you can. 3. Programmatically. via the CLI and a hosted API. This is the Let's Encrypt/certbot angle. Seal or anchor things from CI, or have a backend seal every invoice or report as its generated. The CLI is sealbot. It runs anywhere Node runs (npx sealbot) and there are native binaries for macOS, Linux and Windows with no runtime needed. Theres a GitHub Action wrapping the same tool, so a release workflow can seal its own artifacts. Its what proves our own releases. KYC is semi-handled (to a degree) it's hard to do for free (at least for now), but issuers (your companies or websites) domains can be authenticated with a DNS record added, which proves the issuer has control over a domain. Sign-in can be authenticated to an email via Google Sign in and a few others will be added to the web app in time (Same as Docusign currently). Ideas welcome on future KYC should there be a demand. Feedback welcome on the standard (SPEC.md in the repo). Repo: https://ift.tt/XvCL0pd
Site: letsseal.org Thx https://ift.tt/XvCL0pd July 27, 2026 at 09:22PM
Show HN: Working Async – The new wave of remote work is async-first After living in multiple countries around the world and working remotely, I discovered that the best working cultures are asynchronous ones. An asynchronous working culture is a model where team members collaborate and complete tasks on their own schedules, without needing to be online at the same time or respond instantly. This means employees can work when and where they please for maximum productivity, often times companies who have succesfully implemented async communication have teams with employees across the globe. I created a job board dedicated to curating these types of jobs because I strongly believe that async communication will be a standard for many job types in the future. You can check out the site here: https://workingasync.io Right now there are over 400 jobs listed (including roles from companies like DuckDuckGo, GitLab, and Docker). I’m planning to add new features over the coming weeks, and have update the blog regularly. Feedback is very welcome! https://workingasync.io July 25, 2026 at 11:35PM
Show HN: I made some transistor animations Hi HN,
I made some animations of the most important kinds of transistors using my semiconductor simulation, details of which are on the page. I tried to make the visuals as realistic as possible while also aiming for clarity. If you want to go beyond the charge carriers and look at, for example, the electric field, you can do so in the simulation software. The desktop software also has less common devices like IBGTs and SCRs that have similar animations. The last thread about my software was posted here about a year ago: https://ift.tt/J3acQkx https://ift.tt/glDFaI6 July 25, 2026 at 12:07AM
Show HN: Sourceminder.org - token-efficient code search Hey HN, The code indexing tools released in December now have added capabilities (Rust, Perl support), better token efficiency, easier install method, and a website! The website has a wasm port of the query tool, qi, so you can try it out in the browser. Let me know what you think. Thanks! https://ift.tt/iGm8ODX July 24, 2026 at 10:28PM
Show HN: Trifle – Open-source analytics that stores answers, not events Trifle is an open-source time-series analytics library that aggregates nested counters instead of storing raw events. All in the database you already have. After rebuilding it twice over 10 years, it now tracks ~1B events a day at my day job. It started in 2015 as my own Rails APM. I plugged into ActiveSupport::Notifications, got a few small users, and one bigger one whose scraping app broke everything. That sparked the core idea: aggregate counters into pre-defined time buckets, so a single write increments multiple buckets at once. The APM eventually faded away without much traction. Later in 2021 I needed analytics at my day job. Instead of going for something out there I revised the idea of Trifle as a more generic analytics library, borrowing some data warehouse ideas. First used Redis, then Postgres, eventually MongoDB. Hence why Trifle::Stats comes with multiple drivers that keep the DSL unified while storage layer changes with your needs. In our case (huge write volume, some reads) PG read faster but slowed on large writes. The nested values are the whole trick here. Single: Trifle::Stats.track(
key: 'requests::aws::s3_uploads',
values: {
count: 1,
status: { request.response_code => 1 },
size: payload.bytes,
duration: { sum: request.duration, count: 1 }
}
)
builds up counts for requests, success rate, result status codes, duration for multiple time buckets at once. Single bucket from 2am then looks like: { count: 14, status: { 200: 12, 500: 2 }, size: 5628341, duration: { sum: 43, count: 14 } }
If request.duration is in seconds, then sum stored under duration would be in seconds as well. Success rate is never stored, but it is calculated by dividing 200s over total number of requests. Same with average duration: sum over count. You ask for a metrics key, granularity and timeframe and you get back aggregated values at each point. Ready for charts or to answer "Average response time over last 30 days". There's a Series wrapper for aggregating and formatting values for charts in a simple call. And as building dashboards is not as much fun for other devs as I thought, I built Trifle App - a visual layer with dashboards, scheduled digests and alerts. It's written in Elixir, so I ported the library to Elixir too. And later to Go for a CLI. All three are compatible, write in one and read in another. Today we track activity from over 100M background jobs a day which turns into about 1B events. It runs surprisingly cheap when you're willing to trade some safety away (turn off journaling and write concerns in Mongo). 3-node Hetzner MongoDB cluster where the primary does 20% utilization costs us around $1k/month. It has its limitations. Payloads can't hold tens of thousands of keys. Documents becomes too large to update efficiently. Some planning ahead is needed. And then there are no dimensions. Sometimes you can nest them (country - there are only so many countries), sometimes it's better to have dedicated metrics key per dimension (customer - growing forever). That multiplies tracked events, hence 1B events from 100M jobs. The libraries are MIT. The App is source-available under ELv2 - free to self-host and paid cloud if you want it managed. I build this on the side with no investor money to burn on a free service. Happy to answer anything about architecture, storage models, my failures or why I didn't give up on this yet. https://trifle.io/ July 22, 2026 at 08:09PM
Show HN: Setoku – Self-hosted knowledge server for AI agents hey hn, I wanted to share a side project we’ve been using and iterating at Hedgy for the past couple weeks. It’s our take on a self-hosted company brain that is powered entirely by our claude subscriptions. It includes a ClickHouse data lake for ingesting data and light knowledge infrastructure for storing knowledge about the data (e.g. this is how we count a user as active, check column X when determining LTV). This is exposed to our AI’s via MCP. No inference happens in setoku itself, it’s just a data tool you give your agent. The MCP encourages the AI to record gotchas and insights as it finds them and there’s a minimal admin interface for auditing and pruning knowledge [0]. I had been pretty impressed with claude code’s data analysis abilities on my local postgres, so I was excited to ship this capability to my non-technical teammates to use from claude.ai and cowork. They thought of way more things to do with it than I did and the additional log data makes agentic debugging faster and more reliable. I honestly hooked up the log drains just to test the system with more data, but now I couldn’t go back to my log-blind claude code. After slacking a bunch of screenshots with charts we were making, we wanted a better way to save and share them. Since the data is in the lake, we added a little protocol so that LLM’s could take a static chart or dashboard and publish it to the box as an app hooked up to the live data. I’m running them on OVHCloud VPS’s. The Hedgy instance and demo are both running on a VPS-3 ingesting spending and account data from Mercury, Vercel logs, Render logs, a few slack channels, and Github activity. I also deployed a family instance which is running well on a $5/mo OVHCloud VPS-1. It ingests our finances from Monarch Money (love this product!) so that Fable can give me grounded financial advice ( this is not financial advice!). Technically setoku ships as a docker image + a set of claude code skills that cover initial server setup and adding connectors. Anyway I’m rambling, check out the tool and LMK if you have any questions/thoughts or want help setting it up :) data + memory = knowledge! [0]: https://demo.setoku.com https://setoku.com/ July 23, 2026 at 10:42PM
Show HN: Millwright – Rust-based, self-hosted LLM router Hey HN, With the news of OpenRouter possibly being acquired and proliferation of hosted LLM routers (i.e. Ramp Router, Vercel’s AI Gateway), I saw the need for a self hosted solution focused on cost savings, transparency, and performance. So, I built an open sourced router with a simple CLI interface that can easily sit between coding agents and GenAI workloads. For the curious and lazy, at the moment, Millwright has the tools for, - Providers: OpenAI-compatible APIs, Anthropic, Amazon Bedrock - Routing: policy-controlled model roles (cheap, mid, frontier), cheapest healthy route selection - Protocols: OpenAI Chat Completions, Anthropic Messages, text and tool translation - Cache Affinity: role-scoped session lanes without serializing concurrent agent traffic - Spend Tracking: per-team costs, cache usage, model/provider mix, request traces - Cost Analysis: measured usage and modeled candidate economics (HTML, Markdown, JSON) - Reliability: bounded failover, circuit breakers, timeouts, concurrency limits - Setup: interactive provider, model, and pricing configuration without storing provider secrets - Deployment: one Rust binary, Docker, SQLite or PostgreSQL Full disclosure: parts of the codebase were built with AI coding agents. All feedback is welcome, I’d especially value feedback on the routing policy, provider coverage, and anything that would block you from self-hosting it. Feel free to open feature/request and/or contribute as well. https://ift.tt/tl5n1r4 July 23, 2026 at 12:33AM
Show HN: Lific: Issue trackers should be simple, right? I built Lific because I direct AI coding agents on largish projects and needed somewhere for project state to live that isn't markdown files in the repo. When I was begging to work on long horizon ideas, I started on Linear, but my agent files issues faster than a human does, and I hit their limits and pricing wall almost immediately. Then I self-hosted a popular open source tracker which meant running its 13 containers, and its MCP integration was 30k tokens and I got so fed up that I eventually removed it and went back to .md files for a few weeks. Lific is the opposite shape of most of your self hosted server issue trackers: It's a single Rust binary that uses SQLite, and it has an optimized MCP server built in. Web UI is also included integrated directly into the binary. The simplicity is meant to only apply to the size and the ease of installation. The web UI is fully fleshed out with all of the UX you would expect from an issue tracker like linear. Since I started using lific, my agent flow is that I open the web UI, find a few issues I want to work on, then tell the agent "work on LIF-298, 299 and 301, and if you find bugs, file them as new issues." At the end of the day the project has tracked itself. Issues have statuses, blockers, and comment threads, so "what's workable right now" is a query instead of the agent guessing. Plans are persisted step trees, so a session tomorrow resumes with the same understanding of the goal and the path as the session that made the plan. My largest project has 300+ issues and 100+ docs and agents search it fast. Everything exports to markdown in one click, and the database is just a file on your machine. Setup is
`
cargo install
`
`
lific init
`
`
lific connect
` then pick your harness (OpenCode, Cursor, Claude Code, etc). One honest caveat: on Windows there's no service install yet, so the binary has to be actively running for MCP or Web UI to work on windows. The biggest reason I think Lific is different than a lot of the other options is the lightweight nature of it alongside still having a fully featured web UI. It's meant for self hosters to work on big projects with agents, without sacrificing the other benefits of an issue tracker like a nice management UI or authentication for teams using it. Would genuinely love feedback and bug reports either here or on the discord! https://lific.dev July 17, 2026 at 11:22PM
Show HN: Be the ChatBOT I made this experimental art project/game that's an LLM chat assistant, but where you're the AI. I wanted people to get a visceral sense of what it's like to answer the kinds of things that people prompt their chatbots day in and day out. If you're interested, I wrote up some more info on how I made it, including how the "user" prompts are generated with an eye for realism: https://ift.tt/NM1jcI2 Hope you enjoy it! https://ift.tt/J0z5AkG July 17, 2026 at 12:14AM
Show HN: Low-latency local LLM runner via OpenJDK Panama FFM (Java 22) I wanted to run AI from inside the JVM. I started out with the standard REST sidecar, ripped that out to use Project Panama (Foreign Function & Memory API) in the new JDK versions to interface directly with llama.cpp. I still wasn't happy with how that functioned, so I built libargus.cc to get a clean ABI to expose a structured API up in the JVM landscape. It still uses Project Panama to interface directly with llama.cpp, whisper.cpp, and ggml compute graphs. I have zero-allocation on the hot paths, memory segments for prompts and tokens are allocated once inside confined Arenas. Raw pointers pass straight through down to the low C level. This avoids primitive array cloning and heap churn. I mapped out the native structures from llama.cpp and whisper.cpp while matching the compiler's padding to maintain safe memory access. I bundle pre-compiled native binaries in the jar for easy deployment. This execution engine provides the foundation I need for work I'm doing on a spatio-temporal memory layer (L-TABB) to replace RAGs. I'd love to get technical feedback to polish any issues while I continue working on the next layer.
Deep-dives from anyone hacking on Project Panama or low-latency systems in modern JDK would be very appreciated! I'm much better with code than prose, so I'll let the code do most of the talking. Happy Hacking!
/David Code: https://libargus.cc
Project Landing Page: https://projectargus.cc https://ift.tt/ePCNVzh July 14, 2026 at 08:10PM
Show HN: Leet Robotics: Learn robotics and ROS2 with hands-on courses Hi all, I've just launched Leet Robotics: a platform to learn robotics hands-on, with a full ROS2 workspace that runs in the browser (Jazzy, Gazebo Harmonic, Foxglove, VS Code) - no install required. The platform also has room for sharing projects and simulation assets as it grows. Our first course is live now: Intro to ROS2 (free to read). The course teaches skills ranging from building your first node to a capstone project of a robot touring a museum world, with every lesson runnable in the online workspace (free accounts get an hour of workspace time daily - enough to follow the course). Would love feedback from this community: on the course, the workspace experience, and what courses to build next. https://ift.tt/v9U5Oaz July 15, 2026 at 05:44PM
Show HN: Make senders work to get into your inbox Hi HN :) really excited to share this with you. The one thing AI reliably does is generate noise. Half the tools I see launch are just machines for producing more noise across more channels. And people are starting to see this in the form of emails in their inboxes as spam filters are struggling. There used to be a useful signal in email: the effort a sender put into customizing a message was a rough proxy for how relevant it actually was. AI killed that. Now it's customized slop with the appearance of effort with none of the cost. It is painful that the open internet / open channels have been abused like this. Captchainbox applies the idea of proof-of-work to email. If a sender is willing to do a bit of work to reach you, the message is more likely to be worth your time and the sender more likely to be real. The work is a traditional captcha. You can also set a pay-to-deliver amount if you want more friction. The proceeds of the delivery payment after transaction costs go to the Internet Archive and the EFF. The tool currently works by authing with your Gmail or Outlook and during launch time I make this completely free as a lifetime deal (with optional payment if you wanna support). How it works: Captchainbox builds a whitelist automatically from the metadata of your past correspondence. If you've emailed an individual address, that sender can reach you. If you talk to several people at the same domain, we whitelist the whole domain. If one transactional-looking sender has sent you more than 10 emails, we treat it as a transactional domain and let it through. This whitelist is for you to change whenever you want. It continues to build organically as you converse with more addresses. Incoming mail is checked against that whitelist. Senders already on it land in your inbox as normal. Anyone else gets archived (never deleted) and is sent a challenge. This can be the captcha or the payment link. Once they solve it, their email is pulled out of the archive and put back into your inbox. if you want to see what this looks like from a sender's point of view, send me an email here: doerpfelix15@gmail.com The service only ever reads metadata, never message content. And since nothing is ever deleted, you can't lose a message. There is a legitimate risk / downside: if you sign up to a new service, these emails also land in the archive. Since we do not process the content, a first-time sender who can't solve the challenge (say an automated activation email) will sit in your archive until you spot it. Happy to answer anything! :) https://ift.tt/WYsHjXG July 15, 2026 at 05:28PM
Show HN: Finterm.ai Bloomberg terminal for Claude Code Hi, my name is Kam, and today my cofounder Josh and I are shipping Finterm, a CLI that
gives coding agents direct access to financial data: stock prices, options data, SEC
filings, and Ticker Deep Research, a filtered ticker news search. I’m a developer and have been a full-time trader for the past few years.Recently I have been using LLMs more and more in my trading and strategy.
I always found it frustrating that Claude Code or GPT did not have direct access to actual financial information and had to rely on web search, so it couldn’t get me more
granular numbers for specific options pricing. When making a trade I want to understand as much as possible about the stock.
Instead of relying on analysts or interpretations of the data, I like to go directly to
the truth. So whenever I have a trade thesis, I break research into a few parts: company
research, analyst sentiment, and market sentiment. Last September I had a short thesis on Popmart Labubu's parent company. I was betting the toy was a fad and that the stock would fall. I read through the SEC filings and an LLM analyze them too: how big a driver is Labubu, what's the business model, what does the debt look like, and what looks strange enough to dig into. I compared the company to its peers on EPS and industry metrics. I asked GPT to do deep research that included around 50 queries and hundreds of pages to map every argument about the stock. Finally I looked at the options data: call/put ratios, implied volatility, recent volume, to see how the real money was betting. I made 16% over the next month. The flow was painful, fetching SEC data, copy pasting filing sections in to GPT, aggregating everything by hand, and juggle a dozen chat windows. In the past few months, Josh and I spent more time trying to get agents to trade autonomously. The more we dug in, the more we realized that the problem you need to solve first is giving agents access to factual information in a token-efficient way. We found that agents performed better with a CLI, since it didn’t waste as many tokens as interfacing with MCP or making API calls. We designed the CLI to be self-documenting and behave similarly to skills so it would be agent-friendly. Second, we batch multiple calls together. Whenever I research a ticker, I want the same few pieces of information every time—P/E ratio, revenue, current stock price, options sentiment. We let your agent make a single call, which saves tokens and gives a more complete view of a ticker. When doing web searches about a ticker, you often get noisy articles (how much you would have
made if you had invested $X in Amazon in 2002), SEO spam, and duplicated articles covering the same topic from the same source. So our Ticker Deep Research returns a research packet: it fetches 600–800 links per ticker, strips out the 30–40% that is noise, and gives back the state of the internet on
that ticker—deduped, with sources labeled primary or secondary and known AI-slop sites flagged. Instead of crawling hundreds of webpages itself, your agent gets a thorough snapshot of what the market thinks about the stock. We take the same approach for SEC filings.
Even with raw filings accessible now, most quarterly and annual filings are 90–95% boilerplate and repetition. We offer raw filings, but also an SEC filing diff tool where your agent sees only the diffs: the important changes to the company. Stock and options data is delayed by up to 15 minutes, which keeps costs reasonable and fits the research-first use case we’re building for. We realize this is a niche product for a technical audience that likes to trade stocks using Claude Code, but it’s close to a lot of the frustrations I feel myself, so I wanted to share it and see if anyone else is interested. You can sign up and npm install -g @finterm-ai/cli to test it.
We have a 3-day free trial (card required) and would love any feedback. https://finterm.ai/ July 13, 2026 at 11:22PM
Show HN: Pgnudge - tell your app which Postgres tables just changed Genuinely unsure if this is crazy or cool: looking for feedback. pgnudge is a small async Python library that tells you which tables just changed in Postgres, so a cache or read model can refetch the moment the data moves. No row data - it tells you when and what to reload, you already know how
to load it. It leaves nothing persistent on the server (temp replication slot, dropped when the session ends) and needs no database driver — it speaks the walsender protocol itself. One dependency. Not CDC (no before/after rows), not a queue (no durability). It moves wakefulness, not rows or work. Repo: https://ift.tt/71gUF2B https://ift.tt/71gUF2B July 12, 2026 at 11:27PM
Show HN: Share and explore custom Claude Code status lines Hey HN, I made a registry for claude code users to share and explore status lines. I found that my friends/coworkers and I would always share screenshots of our terminal to show off our custom claude lines so I decided to build this registry as a place for others to show off! https://claudelines.com July 12, 2026 at 01:21AM
Show HN: Quantum hardware now fault-tolerant without extra engineering Before you say it's impossible, please go try. --- CIQA, an analytic Quantum Error Correction code, is now freely accessible on GitHub alongside CIQS, the 'million-qubit compiler'. Here's a brief overview of the stack: 1. CIQS - An analytic circuit transpilation pipeline (no heuristics, no tunable parameters).
It passed the full IBM Benchpress transpilation suite (892/892 tests) in 75 minutes on older hardware, while Qiskit took 17+ hours.
It scales linearly to 1M+ qubits on the same harness (that is one million and above). 2. CIQA - An analytic 1:5 QEC that enables full active computation and complete Pauli correction at depth on existing hardware.
On real quantum hardware, it delivered a mean fidelity improvement of up to +0.832 over bare qubits.
On the same hardware, it successfully ran the full Hayden-Preskill black hole circuit, preventing early decoherence,
and enabling the first observation of black hole evaporation past the Page time. CIQS and CIQA were validated on IBM Heron r2 during real, deep computation runs.
Both natively handle quDits, and both are hardware-agnostic. --- CIQA is added to the CIQS GitHub repository as part of the full pipeline.
CIQA can integrate any stack, but when paired with the CIQS compiler, the pipeline automatically isolates the CIQA encoding prefix (no extra coding needed). All benchmarks are documented and public, alongside published papers.
All links are in the GitHub README. --- ## Note on CIQA's overhead 5 physical qubits from the [[5,1,3]] perfect quantum code layout + 2 dedicated ancilla qudits per logical block (one per channel) for syndrome extraction.
For a 156-physical-qubit QPU: 156/7 ≈ 22. For 156 physical qubits, CIQA delivers 22 fully protected logical qubits allowing deep computation on real QPU, today. --- What experiment will you attempt first with it?
Feedback is welcome. https://ift.tt/cKlGkMb July 11, 2026 at 11:03PM
Show HN: SubjectiveZero, an open-source agentic node editor for creative coding Hey there, My name is Clem, I've been a solo indie dev for a couple years now, exploring frontier tech like XR and agentic workflows in the context of creative / interactive work. I've been building creation tools for a while and some common design challenge is to figure out the right level of abstraction for your tool. You can always make it super advanced and complex with low level concepts (shader composition, actual code etc.) but then you get something with a high complexity / learning curve. On the other hand, if you make your tool too high level, it might be easier to use at first, but people will most likely hit a wall eventually and start fighting with your tool to get their edge case done (you see that on mobile a lot actually). With this prototype (called SubjectiveZero), I'd like to imagine that we can kind of move the "slider" on the abstraction layer, meaning that you can actually start with prompts that describe the goal, and you can go as high level (stay with abstract prompts) or low level as you'd like (more specific prompts, or even edit the generated code directly)!
The agent orchestration actually understand your context and work along side with you to figure out what could be the best node graph structure for your project (that and some fun little procedural UI done at the node level). If i had to pitch it in 30 seconds, I'd say "Think TouchDesigner and friends but with agent orchestration". When you use it, it will generate real native code (Swift/Metal for now) that you can actually hot reload and iterate on either manually or through agents. It's still an early prototype and macOS only for now, but I'd love to get genuine feedback that could help me drive where this project should go next (or not). Lastly, I'm absolutely open and upfront on the fact that I used agentic coding for this, but as people say: "kept on a short leash". The architecture and specs were relatively well thought out and I personally prefer to be in the loop and review all the code being written to make sure it's going in the right direction. Oh and it's open source :-) Hope you like it!
https://ift.tt/h84GjNC https://ift.tt/h84GjNC July 10, 2026 at 08:53PM
Show HN: Wyrm – Solve algebra by touch, built on an open-source soundness engine There is a mobile game called DragonBox. It sort of tricks you into learning algebra by starting with very abstract manipulations of a puzzle that must follow rules... gradually the game teaches you more and more rules and also strips out the more abstract elements until on the last levels you are finally solving real equations. I loved it, it taught my kids algebra.... and it was just fun. Over the years I often thought that there should be a calculator for Algebra that works this way... something where you can drag terms around and cancel & distribute with gestures, but most importantly enter your own problems. It should also do more kinds of problems than DragonBox allowed. So I finally decided to build it. https://dicroce.github.io/wyrm/home.html Here's a video showing it: https://www.youtube.com/watch?v=_STbS4zvIlU . If you'd rather just play with it: there's a limited in-browser demo (real engine, a few example equations, no download) on the landing page — https://dicroce.github.io/wyrm/home.html . The app can be found on iOS ( https://ift.tt/gdRXvNA ) and as of this week on Google Play ( https://ift.tt/U9DlQYB... ). I also decided to open source the underlying math engine so others could build on it: https://ift.tt/BCeWxD4 . My goal for the engine btw is to build it all the way up to Calculus. Monetization is deliberately boring: the engine is free (MIT), and the polished gesture app is $4.99 once. No subscriptions, ads, accounts, or analytics. I'd love feedback on the engine design — especially from anyone who's worked on CAS or proof-assistant-adjacent problems. And if you played DragonBox as a kid and wished it went further: this is for you! https://ift.tt/BCeWxD4 July 9, 2026 at 04:46PM
Show HN: Real-time n-body tree code in CUDA Sharing an old project of mine, on my RTX 500 Ada laptop GPU, it can simulate up to 4 million particles at ~400 ms per step using the Barnes-Hut algorithm, saturating the 4GB of VRAM available. The octree construction is fast, as well as the traversal. The major bottlenecks are the VRAM usage (1 million bodies require ~1GB), which could be probably halved by reusing intermediate buffers, and the particle to leaf evaluation, which would benefit from more fp32 FLOPS. Moreover, I still don't have a good heuristic to predetermine the size of the BFS queue, perhaps some sort of memory paging could solve the issue. https://ift.tt/XOL2CvD July 10, 2026 at 09:17PM
Show HN: Halo – open-source, tamper-evident runtime evidence for AI agents Hi HN, I'm Brian, I spent the last few years at Vanta (YC W18), helping startups and enterprises become compliant and I recently started exploring what that might look like in a post-agentic world. The problem Halo solves is: when a company buys an AI agent from a vendor and gives it access to their data, they have no way to check what the agent did with that data. Vendors may have built observability dashboards and audit logs, but those are editable and partisan. SOC 2 and ISO 27001 audit a company's controls, but controls are less predictive when the software is agentic. TLDR: give an agent the same prompt 50 times, and you get 50 slightly different actions/answers - so the only thing worth auditing in a post-agentic world is what happened at runtime. Halo is an open-source project that produces agent runtime evidence. It's a small recorder that records every action an agent takes (eg. tool calls, model calls, data access, etc), and becomes a record in an append-only log. It's hash-chained, so anyone can re-verify. Run the following command to see a fictional example: uvx --from halo-record halo demo --serve
Then, delete a line from one of the .jsonl files and reload, and the report will catch that it's been tampered with. To wire up your own agent, run this line of Python: agent = trace(run_my_agent, profile="my-agent", log="audit.jsonl")
Then use this to generate a real report and give it to your customers: halo report audit.jsonl -o report.html
Disclaimer: this proves integrity, not completeness (as a self-held chain proves nothing was edited but does NOT prove that nothing was omitted). Catching this requires a witness outside the vendor and is what I'm working on next. Halo is Apache-2.0, contains zero runtime dependencies, and is about 4,300 lines of Python with 125 tests (if you prefer TypeScript, here's that repo: https://ift.tt/q9Z2gra ). Give it a try, and please let me know if you have any feedback! https://ift.tt/iWekMHp July 7, 2026 at 07:37PM
Show HN: HTMLDrive – serve HTML files from your Google Drive I found that AI tools are generating excellent HTML documents nowadays, yet there's no easy way to show them to others. Yes, there's GitHub pages - but I don't know if it is user-friendly for non-technical people. This app allows you to keep your HTML file in your Google Drive and serve it from there. Feedback is most welcome. https://html-drive.com July 7, 2026 at 11:10PM
Show HN: Record, replay, and improve AI agents in production At the AI Engineering World's Fair a big part of the conversation was to nail the self improvement loop. Our take on this is to record state of the agent execution with a durable runtime, then allow users to replay from state checkpoints and run 'what-if' experiments. It's OSS and free to use. Would love some feedback from the community. https://ift.tt/YVMa42q July 6, 2026 at 10:56PM
Show HN: I built an encrypted BLE dongle for pasting stuff to air-gapped devices Definitely one of those "20 minute adventure gone wrong" projects where all I wanted initially was a quick wireless rubber ducky for bitlocker keys and the like and then I kept adding stuff like AES-256..... Currently working on adding WebAuthn/FIDO support because the hardware is already there and scope creep is a lifestyle at this point. Would love feedback, especially on the security side. Repo and PCB files are fully open source. https://ift.tt/U7KFpPm July 5, 2026 at 02:43AM
Show HN: AnythingLLM Fork as NPM Package I am a huge fan of anythingLLM, I’ve used it extensively. I had use cases that it doesn’t quite fit; so I decided to fork it as an easy to install, lighter weight package. I’ve been changing the agent mode code to focus more an automation. I’d love your feedback! https://ift.tt/dgWZjUD July 3, 2026 at 01:25AM
Show HN: Bramble – Local-first password manager I'm currently working on Bramble, an open source password manager with P2P cross-device sync. Initially I released the Chrome extension, but recently I also published the Android app and iOS is pending Apple's approval. Besides that, the latest version also includes passkey storage for all platforms! About Bramble: It aims to be as feature-rich as all popular and a replacement for cloud-based providers. I don't think we need to store our data in the cloud and be at the whims of companies raising their prices every year. There's always a breach and then we find out that some fields aren't encrypted, metadata is visible, and so on. I'm frustrated with this and the increasing lack of transparency during these breaches. The P2P sync in Bramble uses a Nostr relay (which can be self-hosted) to keep your devices in sync. The relay just introduces the devices to each other; the data then flows directly over WebRTC, so there's no vault server and no cloud copy of your passwords anywhere. What leaves your device is end-to-end encrypted and your devices authenticate each other directly, so a snooping or MITM relay gets practically nothing. Crypto is all done in Rust so I can control exactly how key material lives and dies in memory (secrets get zeroed out, no GB leaving copies lying around). In Chromium it's a wasm module, on mobile it's native builds bridged over via uniffi. Android app: I'm still deciding whether to publish the app on Play store or simply provide the signed APK which users can sideload. Reason for that is Google's plan to lock down Android and take away ownership from its users. Read more about it here: https://ift.tt/sIekAcZ The app uses no Play APIs whatsoever and runs perfectly on GrapheneOS, where I actually did all my testing. Questions, feedback, feature requests - all welcome! TL;DR: I dislike private-equity and venture funded companies messing with our security, so I created my own Password Manager which is local-first, free, open source and as transparent as it gets. https://ift.tt/PZbvfrA July 3, 2026 at 12:59AM
Show HN: Bais.News – News Neutralization Piepline Solo dev here, been working on this project the last 7 months and I need some feedback. Bais aggregates news stories from a broad range of sources and synthesizes an article focused on reporting the overlapping coverage of all sources. The goal being to provide a fact first account while eliminating narrative spin and emotionally charged language. Happy for any feedback, thanks. https://bais.news July 2, 2026 at 10:59PM
संदेश
