AI Agents & Automation
-
ACC compiles AI agent trajectories into long-context training data — turning agent runs into a learning signal
ACC (Agent Context Compilation) tackles a clean idea: when agents solve problems, they produce massive trajectories — tool calls, environment observations, dozens of turns. That data is usually thrown away after the task. ACC converts it into long-context training material instead. ## How it works When an agent works a search, software-engineering, or database-querying task,… Continue reading
-
Kanbots runs a Claude Code or Codex agent on every Kanban card — up to 4 in parallel, fully local
Kanbots is an open-source desktop Kanban board where each card is an AI coding agent. Drop a task on a card, and a Claude Code or Codex agent works it. The autopilot mode runs up to 4 agents in parallel, splitting work and evolving the backlog until tasks complete or budget caps hit. MIT licensed,… Continue reading
-
DCP gives AI agents permission, not keys — credentials stay on your laptop, you approve from your phone
DCP is a security layer for AI agents built on a simple inversion: agents get permission, not keys. Your Solana wallets and API credentials stay on your laptop. When an agent needs to act, it asks for permission, and you tap once on your phone to approve. Local-first, open source, works with Claude, Cursor, OpenClaw,… Continue reading
-
buildpipe lets you compose, run, and automate multi-step AI development workflows
buildpipe launched on Product Hunt — a tool for composing, running, and automating multi-step AI development workflows. The idea: chain the discrete steps of an AI build (data prep, model calls, tool invocations, evaluation) into a repeatable pipeline instead of re-running them by hand each time. ## Where it fits This is the same conceptual… Continue reading
-
π-Bench finds proactive assistance still stumps frontier agents — finishing a task is not the same as reducing your burden
π-Bench is a new benchmark testing something most agent evaluations skip: can an AI assistant anticipate what you need before you spell it out? It comprises 100 multi-turn tasks across 5 domain-specific user personas, and the headline finding is sobering — proactive assistance remains hard for frontier agents. ## What it tests Users rarely state… Continue reading
-
Slideshot’s AI agent drives your web app via MCP and returns a polished demo video
Slideshot turns product demo video creation into an agent task. You describe the flow you want to show; the agent drives your web app autonomously through MCP, records the walkthrough, and returns a polished video plus GIF — complete with zoom effects, cursor motion, and a custom intro animation. ## How it works Slideshot connects… Continue reading
-
InstaVM gives AI agents Firecracker microVMs that boot in under 200ms with pause, snapshot, and clone
InstaVM is a production control plane for AI agents — it runs agents like production servers: isolated, observable, and controlled. The core is Firecracker microVMs with sub-200ms boot times, so an agent can spin up a fresh isolated computer faster than most API calls return. ## What the control plane gives you Pause a running… Continue reading
-
Tycono turns multi-agent AI into a tycoon game where you define a company in code
Tycono is an open-source platform that turns multi-agent AI into something that feels like running a tycoon game. Instead of prompting a single agent, you define an entire company — roles, authority, knowledge — in files, and let the agents work together. Launch with `npx tycono`. ## Company-as-Code The core idea: roles and workflows live… Continue reading
-
Qwen3.7-Max ran 35 hours and called 1,000+ tools to write a kernel 10x faster than the vendor code
Alibaba released Qwen3.7-Max on May 19, unveiling it at the 2026 Alibaba Cloud Summit. It’s a reasoning model engineered for long, multi-stage agentic projects rather than short chat — and the headline demo backs that up: it ran 35 hours uninterrupted, called over 1,000 different tools, and wrote an optimized compute kernel that ran 10x… Continue reading
-
Forge guardrails take an 8B self-hosted model from 53% to 99% on agentic tasks
Forge is an open-source guardrail framework from antoinezambelli that takes an 8B self-hosted model from 53% to 99% accuracy on agentic workflows — within 1 percentage point of frontier APIs running the same framework. It showed up on Show HN this week and was presented as a demo at CAIS 2026. ## What’s in the… Continue reading
Top AI Product 