Recurse is a serverless platform for building, testing, and deploying specialist AI agents when generalist models fall short. It centers on a manifest (agent.yaml) that acts as a strict contract: apiVersion and kind, a prompt file, a typed input schema (example: query string, speedup number default 2, task default "Optimize {{ input.query }}."), an output schema that requires a result string, and a tools block (source: tools.py, built_in true, register: search, verify). The manifest gates what the model may accept, call, and claim, validating inputs, applying defaults, and guaranteeing outputs that match the declared schema. Only registered, relevant functions become callable tools, enforcing a minimal, auditable surface for each specialist.
Development runs as an automated design loop: a coding agent generates candidate specialists, runs them through measurements and validators, revises prompts or toolsets, and keeps only candidates that meet verifiable criteria. Once a candidate passes, the manifest is pinned and the specialist can be deployed as a tool, MCP, or bot via CLI (recurse run / deploy), with metrics tracking and adaptive execution built in. Use cases include level design pipelines and RNA sequence design where repeatable, measurable results matter. The platform supports multiple model backends, offers $5 credit for new accounts, is backed by several VCs, and is built by engineers from Stanford, Meta, and other firms.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
AI shopping agents can browse websites, compare products, and make purchases on behalf of users, promising convenience and accessibility. However, their trustworthiness raises concerns about security and privacy risks. (f-secure.com)
Allan Riordan Boll developed a Jev-like wrapper that enables querying both text and vision models by reading token probabilities, allowing flexible and quick assessments. The approach supports real-time webcam analysis by sending images as base64-encoded attachments and retrieving model responses, including visual scene understanding. (allanrbo.blogspot.com)
Large language models (LLMs) are transforming software development, prompting questions about the source of truth in coding. LLMs can work with both high-level frameworks like Rails and low-level code like C, raising issues of trust and editing scope. (intertwingly.net)
A new font design ensures that every token generated by large language models has the same visual width. This approach aims to improve readability and consistency in text display for AI-generated content. (ampdot.mesh.host)
Tesla workers are resisting training the Optimus humanoid robots as replacements, citing difficulties with complex robot hands and job security concerns. Tesla aims to produce 1,000 Optimus robots weekly by the end of 2026 despite these challenges. (arstechnica.com)
OpenAI is experiencing issues with Codex, with root cause identification underway. Most systems, including chat completions and fine-tuning, maintain near 100% uptime, but some components are affected. (status.openai.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.