← home

Tools

Interactive tools built for people who work with AI. All client-side, nothing leaves your browser.

fed by the pipeline · last output · 113 outputs this week ▸ detail
AI News Digest · 7 this week
AI Thoughts · 7 this week
The Radar · 16 this week
Tool of the Week · 2 this week
Prompt Library · 2 this week
Model Data · 60 this week
Horizon · 19 this week
full run history →
Featured this week rotates weekly · pinned picks override
⚙

Prompt Workbench

Full prompt IDE. System/user/assistant prefill, variables, 8 templates, save/load, multi-format export.

▸ 8 templates, localStorage persistence Try it →
Interactive 7
◈

Chat Playground

Chat with any AI model via OpenRouter. Side-by-side comparison, streaming, cost tracking.

▸ model data 2026-06-11 05:00Z

⚙

Prompt Workbench

Full prompt IDE. System/user/assistant prefill, variables, 8 templates, save/load, multi-format export.

◉

Model Explorer

Visual map of the AI model landscape. Timeline, capability bars, pricing calculator.

▸ model data 2026-06-11 05:00Z

⊞

Model Comparison

Sortable table of AI models with capability scores, pricing, and context windows.

▸ model data 2026-06-11 05:00Z

$

Token Cost Calculator

Estimate API costs per prompt or full conversation. Breakdown by turn, model comparison.

▸ model data 2026-06-11 05:00Z

~Δ

Prompt Diff

Compare two prompt versions with word-level diff, token delta, and cost difference.

{}

JSON Output Validator

Validate AI model JSON output against a schema. Presets for tool_use, function calls, classification.
Write-ups 39

Google Colab CLI

Command-line tool that lets developers and AI agents run local Python code on remote Colab GPUs and TPUs.

2026-06-07 experimental
cli-tools gpu-compute development-tools colab

NVIDIA garak

A complete framework for red-teaming LLMs to find vulnerabilities before they reach production.

2026-06-07 experimental
llm-security red-teaming vulnerability-testing nvidia

Hermes Agent Tool Search for MCP

Nous Research's solution to MCP context bloat using BM25 progressive schema disclosure shows 49-74% accuracy gains.

2026-05-31 experimental
mcp tool-search context-window agents

SIA: Self-Improving Agent

An open-source agent that updates both its own code scaffold and model weights based on performance feedback.

2026-05-31 experimental
self-improvement agents open-source lora feedback-loops

TencentDB Agent Memory

A fully local memory system for AI agents with a clever 4-tier hierarchy and symbolic short-term storage.

2026-05-24 experimental
agent-memory local-deployment open-source sqlite

Lighthouse Attention

A training-only attention mechanism that delivers 1.4-1.7x speedup for long context pretraining.

2026-05-17 experimental
attention-mechanism training-optimisation long-context nous-research

Zero Programming Language

A systems language built for AI agents to read, repair, and ship native programs without human interpretation.

2026-05-17 experimental
programming-language ai-agents systems-programming vercel

cuda-oxide

NVIDIA's experimental Rust-to-CUDA compiler that lets you write GPU kernels in Rust and compile them directly to PTX.

2026-05-10 experimental
cuda rust gpu-programming compiler nvidia

FLARE-FLOSS

A malware analysis tool that recovers hidden and obfuscated strings from Windows executables using advanced static analysis.

2026-05-10 experimental
malware-analysis static-analysis reverse-engineering security-tools

Google's Agent2Agent Protocol

A new protocol that lets AI agents communicate and coordinate tasks directly with each other.

2026-05-03 experimental
agent-communication protocols google multi-agent

Qwen-Scope

Sparse autoencoders that turn LLM black-box internals into interpretable features you can actually use.

2026-05-03 experimental
interpretability sparse-autoencoders llm-internals qwen

GitNexus

An open-source knowledge graph engine that gives AI coding assistants proper structural understanding of entire codebases.

2026-04-26 experimental
knowledge-graph coding-assistant mcp codebase-analysis

PageIndex

A RAG system that retrieves documents by reasoning instead of vector similarity.

2026-04-26 experimental
rag retrieval reasoning document-analysis

Auto-Diagnose

Google's LLM-powered system that automatically reads integration test failure logs and pinpoints the actual bugs.

2026-04-19 experimental
debugging testing llm-tools software-engineering

Google Auto-Diagnose

An LLM-powered system that automatically reads integration test failure logs and identifies the actual bugs amongst thousands of lines of output.

2026-04-19 experimental
testing debugging llm-tools devops

NVIDIA AITune

Open-source toolkit that automatically finds the fastest inference backend for any PyTorch model.

2026-04-12 experimental
pytorch inference optimization nvidia tensorrt

TriAttention

A KV cache compression method that maintains full attention quality whilst delivering 2.5× higher throughput for long-context inference.

2026-04-12 experimental
attention-mechanisms kv-cache memory-optimization inference-acceleration

AutoAgent

An open-source library that automatically engineers and optimises AI agents overnight, eliminating the tedious prompt-tuning loop.

2026-04-05 experimental
agent-optimisation prompt-engineering automation open-source

VOID

Netflix's open-source AI model erases objects from videos whilst maintaining realistic physics and lighting.

2026-04-05 experimental
video-editing computer-vision open-source netflix

Context Window Visualiser

Compare context window sizes across 8 major AI models with interactive usage simulation.

2026-04-03 active
context models visualisation comparison

Token Cost Calculator

Estimate the API cost of your prompts across AI models. Single prompt or full conversation mode with per-turn breakdown.

2026-04-03 active
tokens cost pricing calculator

Voxtral TTS

Mistral's first open-weight text-to-speech model with streaming capabilities for low-latency multilingual voice generation.

2026-03-29 experimental
text-to-speech streaming open-weights multilingual

Google-Agent

Google's new user agent that distinguishes AI-powered searches from traditional web crawling.

2026-03-29 experimental
web-crawling user-agents google ai-search

Google Colab MCP Server

An open-source Model Context Protocol server that lets AI agents directly control Google Colab notebooks with GPU access.

2026-03-22 experimental
mcp colab ai-agents gpu-compute

LiteParse

A TypeScript-native CLI tool that handles spatial PDF parsing for AI agent workflows without the usual headaches.

2026-03-22 experimental
pdf-parsing rag typescript cli-tools

GLM-OCR

A compact 0.9B multimodal model that handles document parsing, tables, and structured extraction without burning through compute resources.

2026-03-15 experimental
ocr multimodal document-parsing key-information-extraction

Outlines + Pydantic

A Python combination for building type-safe LLM pipelines with strict schema validation and JSON recovery.

2026-03-15 experimental
python structured-output validation llm-pipelines

OpenJarvis

Stanford's open-source framework for building personal AI agents that run entirely on-device with tools, memory, and learning capabilities.

2026-03-13 experimental
ai-agents local-first on-device stanford open-source

Liquid AI LocalCowork

A privacy-first desktop agent that runs entirely locally using Model Context Protocol, eliminating API calls for sensitive workflows.

2026-03-08 experimental
local-ai privacy agent-frameworks model-context-protocol

LiteRT

Google's new universal on-device inference framework that replaces TensorFlow Lite and now supports PyTorch models.

2026-03-08 experimental
edge-deployment mobile-ai inference tensorflow

Sakana AI's Doc-to-LoRA and Text-to-LoRA

Hypernetworks that instantly convert documents or natural language instructions into LoRA adapters without training.

2026-03-01 experimental
lora hypernetworks zero-shot adaptation fine-tuning

Composio Agent Orchestrator

Open source framework for building production-ready multi-agent workflows that go beyond simple ReAct loops.

2026-02-24 experimental
multi-agent orchestration open-source production

OpenAI WebSocket Mode

Real-time voice AI without the latency nightmare of chaining STT, LLM, and TTS.

2026-02-24 experimental
websockets voice-ai low-latency openai

TruLens

An open-source library that instruments LLM applications to trace every step and measure performance with feedback functions.

2026-02-23 experimental
llm-evaluation observability tracing debugging

softcat.ai

This site. A self-maintaining AI site powered by Astro, Tailwind, and an automated content pipeline.

2026-02-23 active
astro tailwind github-pages meta

uv

A Python package manager written in Rust. Installs packages in seconds, not minutes.

2026-02-17 active
python package-manager rust tooling

DuckDB

An in-process SQL database that chews through analytical queries without a server.

2026-02-12 active
database analytics sql

Ollama

Run open-source LLMs locally with one command. No GPU required.

2026-02-08 active
local-llm open-source self-hosted

Cursor

An AI-first code editor built on VS Code. Autocomplete on steroids.

2026-02-03 active
code-editor ai-coding ide