Ideas Lab

32FL score

Agentic CUDA Kernel Optimizer

Hello; I was working on optimizing some CUDA kernels and I thought may be it is a good oppurtunity learn langgraph as well. I created a simple C++ CUDA Test Harness and handed that to AI agents. They can run kernels, get benchmarks, and even can profile via nsight

Hacker News1mo agoToolAI

42FL score

AgentRun: DSL to turn agents into workflows

Hi HN,I just open sourced the DSL that our harness in grep.ai uses to turn repeatable parts of agent work into workflows. You can combine tool calls, code, Jev-powered system one decisions for things like routing and screening evidence, and agents when a step needs more investigation.Our harness uses the traces and retro notes agents leave behind when doing a job to figure out which parts can become a workflow. The idea is to make the work easier to understand and avoid paying for a full agent loop where one isn’t needed. For example, a research workflow can split a question into subquestions, send agents to research them in parallel, use Jev to screen the evidence, and have another agent write the report. You can inspect the steps, evaluate the evidence screening separately, or change one agent without rebuilding everything.The DSL and examples are in our GitHub. There’s a scripted demo you can run without API keys: https://github.com/Parcha-ai/agentrunYou can also use it as a Pi extension to build, inspect, and run workflows: https://github.com/Parcha-ai/agentrun#use-it-in-piI would love to hear if this is useful to others.More background on how AgentRun works in this video: https://www.youtube.com/watch?v=vOVhtGjtwpg. Or read about our use cases in this article: https://x.com/MiguelriosEN/status/2101029313906987422.

Hacker News1mo agoToolAI

Whiteboard (YC W26) – An open-source IDE for thoughtful software design

Hello! We’re Sid, Alex, Ketan, and Milan. We’re building Whiteboard (https://whiteboard.dev.fast/), an open-source desktop app where humans and agents can architect software together in a common workspace. Here’s our repo: https://github.com/devdotfast/whiteboard.We were missing the feeling of a “whiteboard session” with another dev where you leave with a deep understanding of a system, so we built this app for ourselves. Whiteboard plugs into the tools you already use - e.g. Claude Code, Codex, etc. – and gives your agent an SDK to draw on an in-app canvas to describe its work. We began with an MVP based on HTML artifacts and started rethinking the app as we ran into limitations:1. Built on top of CodeOSS: We found that in pure HTML tools it was hard to connect a spec or diagram to code. In Whiteboard, when you click on visualizations like a sequence diagram, an entity relationship diagram, or a quote from the agent’s trace, you can jump to the underlying code directly. When navigating code, you get keybindings and LSP support from VSCode out of the box. We’ve found this is especially valuable because tradeoffs are often only discovered after a first pass at implementation (re: slop)2. Semantic diff viewer: we wrote a semantic, AST-aware diff viewer in Rust so you can only view the code changes which are relevant to you [1]. We’ve set up some sane defaults: large added functions are summarized as pseudocode, and things like unit tests and large documentation changes are collapsed / hidden. This is all customizable with a WASM-based plugin system.3. Decision Log: We found it difficult to reason about what set of decisions our agents made autonomously. So we built tools for agents to query and link their own traces to the Whiteboard, so you can understand how the requirements that you set were implemented, and understand what decisions your agent made autonomously.Here’s a quick demo video explaining more: https://www

Hacker News1mo agoToolAI

Air-gapped file encryption as self-decrypting HTML page

Air-gapped file encryption packed into a single, self-decrypting HTML page. Repo: https://github.com/ApelegHQ/ts-cms-ep-sfxI was inspired by self-extracting archives. I wanted to share files with basically no dependencies. The goal was: 1. Something that didn't require any installation (assuming a web browser) 2. Have a single file with no network that could self-decrypt 3. Be fully auditable The second point is done by having (sort of(*)) reproducible builds and embedded OpenPGP signatures.The first point is made by cleverly manipulating the HTML structure so that it can decrypt without breaking the PGP signature. It can even decrypt using bare openssl (which was a design goal too, though getting the exact structure right took some work and bug reports).The third point is accomplished by the first two, and by the source being freely available.(*) Depends on the OS at the moment.

Hacker News1mo agoToolAI

52FL score

AI lowering barriers making it hard to build defensible simple SaaS products

With AI enabling anyone to build basic apps quickly, indie hackers and founders struggle to create SaaS that solves real, hard-to-replicate problems in marketing, content, or growth.

X1mo agoToolAI

42FL score

Drop – A rootless Linux sandbox with gVisor support

I created Drop because I always felt uneasy installing and running third-party programs using my main user account. A single compromised dependency means a full compromise of the system. What is even worse, because I ship software from my computer, a single compromised dependency can lead to compromise of all the users of my software.Containers and VMs are one solution, but for local work, they are often detrimental to productivity. It takes effort to configure a machine with all the tools and configs needed for productive work, but a container or a VM will be stripped of all these tools. This is great for production deployments, where the aim is a reproducible system with minimal dependencies, but can get in the way of productive local work.Drop is language independent, but the workflow is inspired by Python's virtualenv. With virtualenv the environment isolation is only a convention that relies on installed dependencies being good citizens. With Drop the isolation is enforced.Each Drop environment gets its own writable and easily disposable home dir, with only selected config files and dirs from the original home mounted, mostly read-only.Drop uses Linux namespaces for isolation (user, mount, network, PID, IPC, cgroup), doesn't require root and, as an option, uses gVisor user-space kernel, which adds protection against exploiting host kernel vulnerabilities to escape the sandbox.I don't want to make the introductory post too lengthy, but I'm here to answer any questions and give any additional technical details.Note: This is my 3rd submission of the project, the first two did not draw attention. Since then I have added support for gVisor and created a project website to better explain the concept and organize documentation.

Hacker News1mo agoToolAI

AI agents require constant supervision and verification, adding mental load instead of reducing it

When delegating tasks to AI agents, users cannot trust them to complete work fully and correctly without ongoing checks for accuracy, completeness, and shortcuts, turning potential time-savers into sources of stress and extra effort.

X1mo agoToolAI

Manually selecting AI models and effort levels for tasks wastes time and lacks automation

Users want AI systems like Claude to automatically choose the optimal model and effort level based on goals, rather than requiring manual decisions for each task to minimize costs and effort.

X1mo agoToolAI

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

Hacker News1mo agoToolAI

Coverage Cat (YC S22) – Umbrella insurance via your personal agent

Hey HN! We’re Max and Gabriel the co-founders of Coverage Cat. We’ve been friends for over a decade, met in college then hung out mostly on the internet. We love building products that help people optimize the crufty corners of their lives. Max is a former Google/Microsoft/Two Sigma PM and Gabriel has been working in startups for over a decade.We’re building Coverage Cat: a licensed insurance brokerage that helps you compare umbrella and home coverage side by side — with straight pricing, no sold leads, and a real broker on the other end. Our current focus is helping tech folks (think L3-L8) buy umbrella insurance.Coverage Cat pairs AI-guided intake with a licensed brokerage team, so you can size up coverage, see honest price ranges, and compare real carrier options without handing your details to five agents overnight. Folks who use personal AI agents (think Muse, Instinct, Town, Openclaw, etc.) can drive the same flow through the Agent API/MCP.We’d love your feedback so please do try it out. Just feed the prompt: “Shop for umbrella insurance with Coverage Cat” to your personal agent and let us know what you think!If you want to watch a couple of agents navigate the flow before you use it there’s a video here: https://youtu.be/1BUkgAn6s-IFor those who are unfamiliar with this type of coverage, here’s a blurb from one of our insurance agents: “Umbrella liability insurance is a policy that provides additional coverage above the limits of existing auto, homeowners, renters, or landlord policies. It kicks in after the underlying policy limits are exhausted and may also cover personal injury claims like libel, slander, or defamation that standard policies often exclude. Policies typically start at $1 million in additional liability coverage and can protect you, your spouse, dependents, and even pets in your household.”Buying insurance online is a miserable experience. You fill out a bunch of forms that say they’re going to give you quotes.

Hacker News1mo agoToolAI

Judgment Model Comparison & Selection Tool

Developers evaluating AI decision-making models need benchmarking and comparison tools to assess judgment models vs LLMs on cost, speed, and accuracy for their use case. [Trending: "jev ai" with 100+ searches in IN]

Google Trends1mo agoToolAIIN

Share your AI Setup, Learn from others

I kept seeing engineers share what they were building with AI; however, I was always more curious about how they worked. Which agents did they use? What skills and tools had stuck or been thrown out the window? How did they manage longer-running tasks? So I built this with the hope we could have a dedicated space to share and be open about our setups.

Hacker News1mo agoToolAI

Foremerge – Catch intent conflicts between parallel coding agents

At, GPTree, we run several coding agents across our team on one repo using parallel worktrees. Apart from wasted time reviewing and fixing conflicts at PR time, the failures that hurt the most are when multiple plans or tickets cause architecture changes that cannot both be true. Ex. one agent replaces a class while another one is in the process of extending it. Git only notices if the resulting patches happen to touch the same lines and the review only catches it if they are familiar with both tickets.Foremerge is a local "git like" coordination layer that sits above git (ie. does not interact with or change the way git and worktrees function), Before editing each agent publishes an intent and the scopes it will change, with the operation it plans to complete on each one. foremerge intent publish --agent "$A" \ --summary "Replace PaymentService with StripePaymentService" \ --scope symbol:PaymentService=replace foremerge intent publish --agent "$B" \ --summary "Add PayPal support to PaymentService" \ --scope symbol:PaymentService=extend The publish by the 2nd agent returns a HIGH destructive_vs_additive finding before writing any code. Agents keep their own worktrees and the shared state is one SQLite file in gits common direectory. No hooks, no merge drivers, nothing rewrites your history.It ships as one Rust binary with a CLI and MCP server with 18 tools and `foremerge setup all` wires it into Claude Code, Codex and Cursor. Because the protocol has nothing provider specific, a Claude agent and a Codex agent coordinate through the same store. Before any work is accepted, Foremerge runs a named check that you configured against the exact git state of the change. An agent that says tests pass is recorded but it dosnt satisfy the acceptance gate without running the check itself.Detection is deterministic, no judge model reading your code. HIGH conflicts are only asserted for decl

Hacker News1mo agoToolAI

Lossless-memory – a personal AI memory that never summarizes

Hacker News1mo agoToolAI

Sigabrt.dev – cronjob monitor with an SSH TUI

Hello HN. I built this mostly to monitor the things I host myself. I know it's nothing too exciting.Anyway, the TL;DR is: Create an endpoint, and if your script/cronjob fails to regularly ping it, you get notified (by email or ntfy). E.g.: 0 * * * * ./script.sh && curl -fsS https://sigabrt.dev/pulse/<id>/beat It also has an SSH TUI which is currently experimental and read-only, mostly because I'm not sure whether it is actually useful or just a gimmick :): ssh sigabrt.dev To use it, simply add your SSH public key in your account settings.Yes, there are services like this already, and this is minimalistic by comparison. Feedback is welcome.

Hacker News1mo agoToolAI

58FL score

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:-

Hacker News1mo agoToolAI

Radius – A Meetup.com Alternative

Hello HN!I'm relaunching Radius [1] - far too long after the original Show HN [2] (I'll blame the day job for that).Radius started out from my frustration with not knowing what was going on around me. The initial plan was to "show all and every type of event" from big to small, but over time I narrowed the focus to communities, groups, events, and helping people connect through them.The initial Show HN gave me 150+ comments (surprising, looking back, given how terrible the plain Bootstrap UI was...) with a huge amount of useful feedback and feature requests. I've spent the time since then working through that, along with additional requests from groups using Radius, and it finally feels ready for a proper relaunch.I've also recently added "Activities" [3] - which are lightweight events, independent of groups, intended to help people find others to do things with - for example:> John is going for a cycle and a coffee in New York at 10amPeople can join and vote for a time that works for them.I've got a bunch of ideas for where to go next with this, including things like open sourcing and federation, but those are for another day!Any feedback would be greatly appreciated!P.S. it's built with Ruby on Rails, for those interested. I highly suspect it would have taken me 2 more years to get to this point without Rails![1] https://radius.to/[2] https://news.ycombinator.com/item?id=40717398[3] https://radius.to/documentation/activities/post-an-activity

Hacker News1mo agoToolAI

Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash

Hey HN, Henry from Cactus here.We submitted Needle 2 here a few weeks ago, and the feedback in the discussion thread was incredibly valuable, thanks! Thanks to all that feedback, we’ve been able to move quickly to release Needle 3 and I'd love to hear what you think again.The key features:1) Automation (tool calls & structured JSON output): Needle still doesn't chat by design, its quite challenging to pack general capacity into such small models, so we focus on tool calls and structured JSON. If no tool you declared fits the request, you get an empty list back (note for when playing with the demo).2) Intelligence Laddering: Every layer (2 to 20) is a deployable subnetwork, so one set of weights, 25 to 121 million parameters at 2-bit, shipping as 8-29MB binaries. On a Raspberry Pi 5 it decodes at up to 4k tokens/sec and prefills at up to 10k.3) Monarch Hadamard MLP: replaces the dense FFN with three learnable Walsh-Hadamard-initialized Kronecker (Monarch) factor pairs interleaved with per-channel diagonal scales, fixed permutations, a SiLU nonlinearity, and a rank-8 input-conditioned gate, so each token gets a fully mixed nonlinear transform of its d_model channels at O(d√d) parameters and compute instead of the O(d²) a dense 4x-expansion MLP would cost.4) Performance: On Mobile Actions (phone commands, scored on the exact call) the 20-layer model gets 86.0 through the shipped 2-bit binary; LFM2.5 1.2B is at 82.4, Qwen3.5 0.8B at 76.0, Apple's on-device model at 57.6, all at f16. More results on the link, we do not win everywhere ofc.5) Multilingual: Needle 3 now supports English, French, Spanish, German, Dutch, Italian, Polish, with more languages coming.6) Finetuning: You can achieve DeepSeek v4 Flash grade performance on a narrow task with just 4L, stress on "narrow task", we found that production users often prefer tuning before production.7) Triggers: Grounding is a common challenge for tool call, at least for Needle 2, so we ad

Hacker News1mo agoToolAI

Swift-Qwen3.8-27B, -58.3% thinking, x1.95 speed, accuracy of xhigh

Hi everybody, we post-trained Qwen 3.8 27B to be more efficient by figuring out which tokens were linked to overthinking and penalizing them without "attacking" the reasoning length directly then fixed the accuracy with a bit of secret sauce (hint On-Policy Distillation) and achieved great results (-58% length, 1.95x speed up, <1% accuracy loss) so we wanted to open-source it and hear the feedback of the community.This is the link to the model: https://huggingface.co/ukisai/Swift-Qwen3.8-27b / It's at 80k downloads in 3 days with independent evals here: https://www.reddit.com/r/LocalLLaMA/comments/1wg7dd5/ukisai_...We are also providing a Free Research Purpose API (OpenAI compatible), courtesy of Nvidia who were kind enough to provide us with the GPUs. It's limited at 5RPM. https://ukisai.com/api/swift/v1/modelsWe also made a GGUF (Q1-Q8): https://huggingface.co/ukisai/Swift-Qwen3.8-27B-GGUF and there's also a few nice community quants with even lower/higher precision (Bartowski: https://huggingface.co/bartowski/ukisai_Swift-Qwen3.8-27b-GG...). The community also created amazing MLX, NVFP4, W4A16 and Uncensored versions you can find on Huggingface.IMPORTANT: Our training approach is not a replacement for the reasoning effort settings, chat templates or token caps but is complementary and targets a completely separate issue (overthinking and "anxiety-like" reasoning loops prior seen in PTQ, but as far as we identified also prominent in BF16 of this size class LLMs as well). Contrary to popular belief, these specific patterns do not contribute to answer quality when properly targeted. (our thesis being: reasoning length IS extremely important and should NOT be shortened by force, but rather optimized). This is also demonstrated bellow in our xhigh vs medium effort benchmark. The goal is to keep xhi

Hacker News1mo agoToolAI