Favorite Tools

Anthropic

Claude API & Claude Code

Primary AI provider. Claude powers multi-agent systems, RAG pipelines, code generation, and agentic development workflows. The context window and instruction-following quality make it the right choice end to end.

Google

Gemini API & AI Studio

Gemini models for multi-modal tasks, long-context analysis, and cross-validation in multi-agent workflows. AI Studio for rapid prototyping.

OpenAI

GPT & Embeddings

GPT-4 for specific use cases, embedding models for vector search pipelines. The ecosystem integration is mature.

Groq

LPU Inference

Ultra-low-latency inference for speed-sensitive workflows. LPU hardware delivers sub-second responses on open-weight models.

Nvidia

NIM & GPU Infrastructure

NIM microservices and GPU-accelerated inference. The hardware substrate underneath most of the AI stack.

OpenRouter

Model Router

Unified API across dozens of model providers. Route to the best model for each task without managing individual API keys.

DeepSeek

Cross-Model Reviewer

An independent reviewer from a different architecture. When two models grade the same work, a separate lineage surfaces disagreement a single family would miss - the cross-validation the convergence method behind Forcing Consensus depends on. Downloadable weights when a job needs to run on-prem.

Mistral

Cross-Model Reviewer

A second-opinion evaluator from a different lineage, brought in to widen the spread of judgments before reconciliation. Strong multilingual performance, with downloadable weights when a client needs EU-based inference or on-prem compliance.

Claude Managed Agents

Hosted Agent Runtime

The reasoning loop for Limner runs here: Claude does the planning and tool choice on Anthropic's managed platform while the tools execute in isolated Cloudflare Workers. The harnessed-agent pattern without standing up the harness yourself.

LlamaIndex

RAG Framework

The standard framework for building retrieval-augmented generation pipelines. Data connectors, indexing strategies, and query engines purpose-built for connecting LLMs to data.

Vercel AI SDK

Streaming & Tool Use

TypeScript SDK for building AI applications with streaming responses, tool calling, and multi-provider support. The abstraction layer for AI-powered UIs.

Weights & Biases

Experiment Tracking

MLOps platform for tracking experiments, versioning models, and visualizing training runs. The standard for reproducible machine learning development.

Ollama

Local Model Runtime

Run open-weight models locally with a single command. Persistent local inference on commodity hardware - the runtime layer for personal-AI experimentation.

LM Studio

Local Model GUI

Desktop application for discovering, downloading, and running local LLMs. Model evaluation and testing without cloud dependencies.

vLLM

Production Inference Server

High-throughput, memory-efficient inference engine for LLMs. PagedAttention for serving models at scale. The production backend for self-hosted model deployments.

MidJourney

Image Generation

Concept exploration and base imagery in the original Limner: Pixel pipeline - the editorial layer for the first VGA-era asset batches. The aesthetic control and style consistency are unmatched for creative direction.

PixelLab

Pixel Art Generation

Pixel-art translation in the original Limner: Pixel pipeline, purpose-built for the medium rather than adapted from photorealistic models. The current Limner generates through OpenAI gpt-image-1 and Recraft instead.

ReCraft

Design-Grade Generation

Vector and raster generation with strong design sensibility, and a generation backend in the current Limner agent. Useful for UI assets, icons, and illustrations that need to feel designed rather than generated.

Elicit

Research Assistant

AI-powered research tool for finding, analyzing, and synthesizing academic papers. Useful when a project needs literature review before a design decision.

ElevenLabs

Voice & Audio AI

Voice synthesis and audio generation. The quality ceiling for text-to-speech and voice cloning in production applications.

Hugging Face

Model Hub & Inference

Model discovery, hosted inference, and the open-source ML ecosystem. The default starting point for evaluating new models and architectures.

Perplexity

AI Search

Citation-backed AI search for research and fact-checking. Replaces the Google-then-read-ten-tabs workflow with direct answers and sources.

Brave

Privacy-First Browser & Search

Privacy-respecting browser and search API. The Search API powers tool-use pipelines where you need web results without tracking overhead.

AWS

Cloud Platform

S3, Lambda, Bedrock, SageMaker - the enterprise cloud ecosystem. Certified AI Practitioner with hands-on experience across compute, storage, and AI services.

Docker

Containerization

Container runtime for reproducible builds and deployments. The standard abstraction layer between development environments and production infrastructure.

Kubernetes

Container Orchestration

Container orchestration for production workloads. Deployment, scaling, and management of containerized applications across clusters.

Terraform

Infrastructure as Code

Declarative infrastructure management for Cloudflare resources. Version-controlled, reviewable infrastructure changes through HCL configurations.

GitHub Actions

CI/CD Pipelines

Automated build, test, and deployment workflows. Powers the CI/CD pipeline from push to Cloudflare Pages deployment.

Vercel

Frontend & Serverless

Deployment target for Next.js projects outside the Cloudflare stack. It hosts PHINEAS. AI SDK integrations, preview deployments, and edge functions.

Tailscale

Mesh VPN

Zero-config mesh networking across devices. Makes the lab PC, development machines, and cloud services behave like they're on the same LAN.

Wrangler

Cloudflare CLI

Cloudflare's command-line tool for developing and deploying Workers, Pages, D1, R2, and KV. The single entry point for the entire Cloudflare developer platform.

Renovate

Dependency Automation

Automated dependency updates with grouped, scheduled pull requests. Keeps lily-livered and other repos current without manual version bumping.

Lighthouse

Performance & Quality Audits

Performance, accessibility, and best-practice audits in CI. Runs several times per push on lily-livered and takes the median for a stable score.

Firebase / Firestore

App Backend & Datastore

Auth, hosting, and the part-of-speech-aware Firestore vocabulary database behind PHINEAS. The document model carries per-word CEFR levels as data the model looks up rather than guesses.

Neon

Serverless Postgres

Serverless Postgres with branching. Holds the structured grading metadata in Forcing Consensus, so every retrieved passage traces back to its source.

PostgreSQL

Relational Database

The industry-standard relational database. Full SQL, JSON support, extensions ecosystem. The foundation when projects need more than D1 SQLite at the edge.

Redis

In-Memory Data Store

High-performance caching, pub/sub messaging, and session storage. The speed layer for applications that need sub-millisecond data access.

BigQuery

Data Warehouse

Serverless data warehouse for analytics at scale. SQL-based analysis over petabyte-scale datasets without infrastructure management.

Prometheus

Metrics & Monitoring

Pull-based metrics collection and time-series database. The standard data source for Grafana dashboards and alerting rules.

Grafana

Observability Platform

Dashboards and visualization for metrics, logs, and traces. The single pane of glass for monitoring deployed infrastructure and AI pipeline health.

Loki

Log Aggregation

Horizontally scalable log aggregation designed for Grafana. Label-based indexing makes it cost-effective for high-volume log streams.

Sentry

Error Tracking

Real-time error tracking and performance monitoring. Stack traces, breadcrumbs, and release tracking for production applications.

Promptfoo

Prompt Regression Testing

Snapshot prompts, define assertions, catch regressions before they reach production. Open source, runs in CI. The test framework LLM applications were missing.

Inspect AI

Evaluation Harness

UK AI Safety Institute's structured evaluation framework. Designed for capability, safety, and alignment testing with reproducible scoring and dataset versioning.

OpenAI Evals

Evaluation Reference

The canonical reference for LLM evaluation patterns. Templates, registries, and the evaluation vocabulary the rest of the field borrows from.

LangSmith

LLM Observability

Tracing, evaluation, and monitoring for LangChain applications. The observability layer when LangChain is in the loop and you need to see what each chain step is doing.

NVIDIA SkillSpector

Agent Skill Security Scanner

Static security scanning for Claude Agent Skills. Scores a skill 0 to 100 and gates the build on HIGH or CRITICAL findings - the security half of every Skill Card in Skill Foundry.

Qdrant

Vector Database

The vector store behind Forcing Consensus and the Embassy lab. Fast similarity search over conversation archives and document chunks, with payload filtering for hybrid retrieval.

pgvector

Postgres Vector Extension

Vector similarity search as a Postgres extension. The default when embeddings need to live alongside relational data without a separate vector database.

Mem0

Memory-as-a-Service

Hosted persistent memory layer with API-first design. Drops into any agent application without standing up vector DB or retrieval infrastructure yourself.

OpenClaw

Agentic Personal Assistant

Agentic personal AI assistant that takes actions across platforms. Open-source, action-oriented architecture - the AI that actually does things, not just answers questions.

ZeroClaw

Local-First Private AI

Private AI assistant that runs 100% locally. Multi-platform (Telegram, Discord, WhatsApp) with no cloud dependency - your data never leaves your machine.

LobeChat

AI Chat Framework

Open-source chat framework powering multi-agent conversation interfaces. The venue for the Forcing Consensus collaborative synthesis round, where competing evaluators reconcile on the record.

OpenJarvis

On-Device Personal AI

Personal AI that runs on personal devices. Architecture for on-device assistants from Stanford's Scaling Intelligence Lab - privacy and latency both solved at the substrate.

TypingMind

Premium Chat UI

Enhanced chat interface for Claude, GPT, and Gemini. Custom personas, prompt library, and conversation management - the UI layer when the native apps aren't enough.

Python

Language

The lingua franca of AI/ML development. Data pipelines, model training, scripting, and API services. Every AI project touches Python somewhere.

TypeScript

Language

The language underneath the entire stack. Type safety from schema validation to API boundaries to UI components. Every project is TypeScript-first.

Bun

JavaScript Runtime

The runtime and toolchain behind the Forcing Consensus pipeline. Fast installs, native TypeScript, and a single binary in place of the Node-plus-bundler stack.

Astro

Static Site Generator

Content-first framework that sends zero JS by default. Powers this site and most web projects. Island architecture for interactive components without the full bundle tax.

Next.js

React Framework

Full-stack React framework for projects in the Vercel ecosystem. Server components, API routes, and middleware for applications that need SSR or ISR.

Cloudflare

Edge Infrastructure

Workers, Pages, D1, KV - the serverless edge stack behind this site and Limner's tool execution. Single wrangler.toml, one billing dashboard, global deployment with sub-50ms cold starts.

Tailwind CSS v4

CSS Framework

v4's CSS-first config and oklch color system are significant upgrades. The @theme block in global.css replaces the entire config file.

React 19

UI Islands

Astro island components for interactive UI. React 19's concurrent features and improved hydration make the island pattern practical.

Zod / Valibot

Schema Validation

Zod for general validation, Valibot for edge-optimized contexts. Both enforce type safety at runtime boundaries.

pydantic

Python Schema Validation

The single source of truth for the Skill Card schema in Califa. Runtime validation at the boundaries so the prose and the code can't disagree.

Playwright

E2E Testing

Cross-browser end-to-end testing and automation. Clean API that doubles as an agent tool - browser control, scraping, and UI validation in one framework.

Biome

Lint & Format

One fast tool for linting and formatting in place of the ESLint-plus-Prettier stack. A single pass in CI on lily-livered and other projects.

Serena

Semantic Code Navigation

AI-native code navigation that understands symbol relationships, not just text matches. Semantic search across codebases.

BBEdit

Text Editor

The reliable workhorse. Pattern matching, regex, multi-file grep, disk browser. Not glamorous but it's been the right tool for 30 years.

Warp

Terminal

AI-native terminal with built-in command completion and workflow blocks. Session history and shareable runbooks for development workflows.

Obsidian

Knowledge Management

Local-first markdown PKM. Project notes, AI research, architecture decisions. Plain markdown files are the real value - no lock-in.

Affinity Suite

Design & Publishing

Designer, Photo, and Publisher - the full creative suite. Professional-grade design tools without the subscription model. Used for presentations, diagrams, and visual assets.

Discord

Community & Voice

Community channels, voice collaboration, and bot integrations. Where technical communities live - from MidJourney to open-source projects.

Google Workspace

Productivity Suite

Docs, Sheets, Slides, Drive, Meet - the collaboration backbone. Real-time co-editing and deeply integrated search across all content.

NotebookLM

AI Research Notebook

Google's AI-powered research tool. Source-grounded conversation over uploaded documents. Strong for synthesizing multiple papers and reports.

GitHub

Version Control

Source of truth for all projects. Issues and PRs for project coordination. The collaboration layer on top of git.

Asana

Project Management

Task tracking and project coordination across workstreams. Portfolio-level views for managing multiple concurrent projects.

Linear

Issue Tracking

Modern issue tracking built for engineering velocity. Keyboard-first interface, cycles, and project views. The tool engineering teams actually want to use.

Atlassian Suite

Jira, Confluence, Bitbucket

Enterprise project management and documentation. Jira for structured workflows, Confluence for technical documentation, Bitbucket for enterprise git.

Miro

Whiteboarding

Collaborative whiteboard for diagramming, brainstorming, and visual planning. The default surface for architecture discussions and workshop facilitation.

Notion

Documentation & Wikis

Flexible workspace for documentation, databases, and project wikis. The block-based editor handles structured and unstructured content equally well.

Jupyter Notebook

Interactive Computing

Interactive notebooks for data analysis, model experimentation, and documentation. The standard environment for exploratory AI/ML work.