Field guide / primary domain
198
sources in this field
Updated July 16, 2026
Current thesis
The shortest path to orientation.
Developer tools converge on two themes with a third emerging: standardized knowledge formats for agent-accessible wikis (Google's OKF storing markdown directories agents query/edit programmatically, replacing Obsidian/Notion at the knowledge layer), standardized code review practices (Google's eng-practices repo providing bidirectional reviewer/author guidance, LGTM/CL terminology), and designer-level aesthetics in AI-generated frontends (impeccable.style's Codex design skill, imagegen-frontend-web workflows). Tools reduce friction between raw content and structured formats (Mintlify, Defuddle, Google CodeWiki, Tolaria/Cabinet/ByteRover as KB-as-agent-surface) and reinvent infrastructure for AI coding: agent-native primitives becoming first-class (Camofox anti-detection browsers for agent crawl armies, Browserbase skill catalogs for web agents, Astropad Workbench headless Mac monitoring, FieldTheory local bookmark sync, /ss screenshot eyes, Lamina Labs whiteboard animation SDKs, /ultraplan cloud planning, Monitor tool event-driven scripts). Free inference is mainstream (NVIDIA ~80 models). Terminal emulators redesigned for agentic workflows (Ghostty-based with vertical tabs and embedded browsers). UI exploration decoupled from git branching (UIFork). Excalidraw's $0/110K-star whiteboard displaced Miro at Google Cloud, Meta, Notion, Obsidian, HackerRank. Agent-tooling layer consolidated: Pipecat/browser-use/Mem0/Composio/RAGFlow/Dify with Mastra (1.77M monthly npm, YC) as TypeScript-first option; MCP is survival requirement ("no MCP, swiftly cancelled" signals 12-month death timer); reusable site-specific skills are web-agent libraries; free inference mainstream via NVIDIA; multi-agent OSs (NovaStation, Hermes v0.12.0 Kanban) demonstrate AI-native command-center pattern. IDE-terminal frontier rebuilds editing for agents: libghostty terminals with vertical tabs/embedded browsers, Decode's browser+whiteboard in Claude Code, UIFork UI/git decoupling, Astropad Workbench headless Mac eyes. Infra-devex standardizes one-click deployment (OpenClaw, Convex+Vercel), real-time token cost visibility (CodexBar), multi-model/device dev loop (gstack 6-specialist team in 30 seconds, @conductor_build Opus-plan/GPT-5.5-review/Playwright-validate for ~$400/month, Codex plugins kill context switching, always-on Mac Studio/Mini nodes reachable from iPhone/iPad/Mac). Knowledge-tooling densest sub-area: content-to-structured pipelines (Mintlify, Defuddle, brain-ingest, MarkItDown), Obsidian-as-agent-surface (smart-connections + qmd MCPs, Claude skills mapping file-based notes), KB-as-agent-surface (Tolaria's native MCP over plain-markdown vault, Cabinet, ByteRover's unified relevance index), local bookmark graphs (FieldTheory, Siftly), "to draw" as agent primitive (Excalidraw, Lamina Labs, Hyperframes). Automation spans browser/file control (dev-browser, agent-browser at 82% fewer tokens, WebMCP), real-time issue tracking (LogRockets, Symphony assigning Codex agents), productivity skills (/ss, Codex Chronicle), structured /goal prompts with explicit verification, systematic diagnostics (speedtest/DNS/MTU loops), replicable verticals (ml-intern for ML research). Ecosystem: tools rebuilt for agent-first workflows (Core AI Workspace fusing Slack+Linear+Notion), Anthropic platform releases (Monitor, /ultraplan, official setup plugins), hiring signals confirming realignment (OpenAI $280K Forward Deployed Engineers screening for "actual loop" not LeetCode), local-first agentic stacks (JustHireMe's Tauri+FastAPI+SQLite+KuzuDB+LanceDB, Codex-built Superhuman replacements, LibreChat self-hosted wrappers) displacing expensive SaaS. Common thread: tools for human workflows rebuilt/extended for agent-first workflows; sub-ecosystems large enough to warrant own discovery layers; Anthropic-as-Platform increasingly anchors (independent tools slot around, not replace); git-based knowledge systems hit 2.3GB+ scaling walls (forcing SQLite migration); taste/judgment automated (Mintlify documentation best practices, Refero's 2,000 DESIGN.md); local-first AI winning (Defuddle, brain-ingest, dev-browser, claude-smart all run locally).
Evidence board
Claims worth carrying forward01Leaner system prompts improve GPT-5.6 performance: internal coding-agent evals show 10–15% score gain, 41–66% token reduction, and 33–67% cost reduction. Trim repeated rules, no-op examples, and unrelated tools while keeping outcome, success criteria, constraints, and tool-routing rules.
02GPT-5.6 migrates better with an outcome-first structure: state what good looks like, add stopping conditions, and use ALWAYS/NEVER only for true invariants. Decision rules outperform blanket absolutes for judgment calls like when to search, ask, or keep iterating.
03GPT-5.6 is more concise by default than GPT-5.5; broad brevity instructions ('Be concise') may now be unnecessary or counterproductive. Use the text.verbosity parameter (low/medium/high) for consistent cross-request control, reserving the prompt for task-specific length requirements.
04Define autonomy and approval boundaries explicitly for agentic GPT-5.6 tasks: name safe local actions (read files, inspect logs, run tests), require confirmation only for external writes, destructive actions, or scope expansion. Repeating 'ask first' causes unnecessary approval pauses.
05Programmatic Tool Calling (PTC) fits bounded record-reduction stages—filtering, joining, deduplication, batching—not simply parallel or dependent calls. Use direct calls when results require semantic judgment, citations, approval, or when each result may change the next decision.
06When migrating to GPT-5.6, preserve current reasoning effort first, run evals before touching the prompt, then remove scaffolding incrementally. Do not rewrite the full prompt stack at once—you cannot isolate whether regressions came from the model, reasoning setting, prompt, tools, or runtime.
07GPT-5.6 prompt migration checklist: switch model → run evals → trim obsolete scaffolding → make one surgical fix per measured regression → re-run evals. Debug regressions on small real traces; identify the failure mode, find the conflicting instruction, edit, rerun.
08For grounded answers, absence of evidence should not automatically become a factual 'no.' Cite only retrieved sources, label inference separately from supported facts, state conflicts between sources, and narrow the answer or report missing evidence instead of guessing.
Adjacent fields
Latest evidence
Adam.GPT
OpenAI's official prompting guide for GPT-5.6 Sol recommends trimming redundant instructions and tools, framing prompts as outcome definitions with stopping conditions, and defining explicit autonomy boundaries for agentic tasks. Internal evals show leaner system prompts improved scores 10–15% while cutting tokens 41–66% and cost 33–67%.
Jason Zook
Jason Zook shares a two-model software development workflow using Claude Opus 4.7 for planning and building, GPT-5.5 for adversarial review, and Playwright for automated UX/UI testing. The loop runs plan→review→update→build→review→fix→deploy, with ~$400/month replacing what feels like a full dev team.
Vasu-Devs
JustHireMe is an open-source, AGPL-licensed desktop app (Tauri + React + Python/FastAPI) that addresses broken job search UX with a local-first, explainable AI pipeline: scrape → quality gate → rank → match (GraphRAG + vector) → generate tailored application materials. All data stays on-device by default, LLM usage is keyless-capable via Ollama/Claude Code CLI, and the architecture prioritizes human-in-the-loop control over blind automation.
Charlie Hills
gstack is Garry Tan's open-source Claude Code setup that transforms a solo AI assistant into a structured 6-role virtual engineering team (CEO, Eng Manager, Designer, QA, Release Engineer, Doc Engineer) via 23 slash-command skills. Tan claims ~810× his 2013 productivity rate measured in normalized logical lines of code, shipping 3 production services and 40+ features in 60 days part-time. The system enforces a think→plan→build→review→test→ship→reflect sprint discipline where each skill feeds context into the next, and supports 10-15 parallel sprints via Conductor.
Ruari Wood-Freeman
A GitHub-based plugin marketplace pattern allows centralized management of multiple Claude plugins with sub-skills and agents across an entire organization. Updates to the central repo automatically sync to all users' Claude instances, eliminating manual distribution overhead.
Ben Tossell
Ben Tossell cancelled Superhuman and built a custom email client using Codex. It mirrors Superhuman's UX patterns but is fully customisable, runs on Gmail CLI, and is agent-native. The core AI logic is minimal: reading labels/archives and updating Gmail filters.
Charly Wargnier
Tolaria is a free, open-source (AGPL-3.0) desktop app for Mac and Linux that implements Karpathy's LLM wiki concept — a shared knowledge environment for humans and AI agents. It uses plain markdown, Git-backed vaults, and includes a built-in MCP server for Claude Code integration. Built with Tauri, React, and Rust; 100K+ lines, 85% test coverage, 9.9/10 code health.
Ole Lehmann
Ole Lehmann ranks 13 integrations for the Hermes personal agent, with Firecrawl (clean web data, fewer tokens), Browserbase (real browser sessions), Google Workspace, GitHub, and Obsidian as the essential five covering ~80% of needs. Composio handles one-click setup.
Matt Slotnick
In May 2026, both Anthropic ($1.5B JV with Blackstone/Goldman/H&F) and OpenAI ($10B 'The Deployment Company' with TPG) launched field deployment engineering vehicles, signaling that deployment bottlenecks—not model capability—are the limiting factor. Anthropic simultaneously went vertical in financial services (40% of top 50 customers are financial institutions), while enterprise SORs like SAP, Workday, and ServiceNow moved to lock down third-party agent access. Salesforce countered with 'Headless 360,' betting on openness and a partner ecosystem instead.
Nick
Nick Baumann describes a multi-device Codex setup where a Mac mini serves as an always-connected 'home' node and a MacBook becomes a 'satellite,' with mutual SSH and cross-device thread continuity giving him phone-initiated, 24/7 Codex access across all three devices.