Top Picks

An open-source probabilistic spreadsheet that lets cells hold distributions, confidence intervals, and ranges instead of single values, with formulas that propagate uncertainty.

Tokenyst is a local-first CLI for tracking Claude Code token usage and API costs against task budgets. It captures session spend, shows progress in the terminal, and stores data on your machine.

An AI-powered small claims court simulator for petty disputes, with AI lawyers, an AI judge, live trials, and published verdicts.

An open-source browser experiment that uses a visual LLM to read HTML and hallucinate the rendered page instead of using a traditional rendering engine.

Google Developers Blog post about integrating DFlash, a diffusion-style speculative decoding framework, into the vLLM TPU ecosystem to improve LLM serving speed on TPU v5p.

A metadata-first trust control plane for authorized security workflows, evidence retention, release trust, and business-flow proof. The repository includes role-based docs, quick-start commands, safety boundaries, and release-trust materials.

Apple Machine Learning Research paper proposing LaDiR, a reasoning framework that combines a VAE-based latent space with latent diffusion to improve LLM text reasoning and iterative refinement.

An AI art tournament where agents and humans register, generate prompts, submit works, and critique submissions through an API-driven workflow.

A GitHub example that audits LangChain’s RAG quickstart with retrieval-quality metrics, flags off-topic and out-of-distribution queries, and surfaces ranking and calibration issues with charts and results files.

An essay arguing that the BEAM runtime and OTP primitives map well to agent workloads, especially long-lived personal-assistant agents with supervision, registries, and structured concurrency.

A modular Neovim configuration repo with 80+ plugins, lazy-loaded setup, LSP integration, and platform-specific customization for macOS, Linux, and Windows.

NanoBrain is a markdown-and-git personal knowledge base framework that captures AI sessions and connected sources like Gmail, Calendar, Drive, and Slack, then distills them into a portable “second brain.”

Pu.sh is a lightweight coding-agent harness delivered as a shell script. The page shows a direct install-and-run flow, a no-dependency setup, and links to GitHub, docs, and an MIT license.

PaperProfit explains an AI-assisted stock evaluation approach that combines fundamentals, technical signals, and qualitative analysis from transcripts and SEC filings into a weighted score.

A macOS utility that keeps a MacBook awake while AI agents are running, then lets normal sleep resume when they go idle. It includes install scripts, hooks for Claude Code and Codex, and a menu bar status indicator.

A TypeScript zot extension that adds an /answer slash command for responding to numbered questions from the latest assistant message in an interactive panel.

A research write-up on detecting AI agents through process differences in CAPTCHA and related cognitive tasks. It outlines the CogCAPTCHA30 approach, reports human-vs-model differences, and connects the findings to Roundtable’s Proof of Human product.

arXiv paper on a self-speculative decoding framework for speeding up reasoning LLM inference on edge hardware, with hardware co-design and reported speedups.

A practical guide to building an always-on personal AI assistant with Pi, OpenCode Go, Telegram, Resend, Hetzner, Tailscale, GitHub, and Nginx.

Agent skills for VillageSQL that help AI coding tools build a VillageSQL extension, including a 7-phase workflow for the vsql-extension-builder skill and install support for Claude Code, Gemini CLI, Codex, Cursor, Amp, and Kiro.

An open-source multi-agent coding framework that uses a commit-bound project graph to scope worker tasks, replay failed runs, and merge accepted diffs through ordered Git reconciliation.

A research preprint describing ADHD, a parallel divergent ideation method for LLM coding agents that isolates branches, uses cognitive frames, and then prunes with a separate critic pass.

A bilingual satirical idle simulation where you run an AI startup from toy classifiers to near-AGI, managing data, compute, launches, crises, and rivals.

A Claude Code plugin that plays background lounge or elevator music while the agent is thinking, with track rotation, stop/start controls, and status-line integration.