Skip to content

Instantly share code, notes, and snippets.

View unitedideas's full-sized avatar
🏠
Working from home

Shane Cheek unitedideas

🏠
Working from home
29 followers · 1 following
  • Portland
View GitHub Profile
@unitedideas
unitedideas / claude-code-to-codex-local-harness-migration.md
Last active May 24, 2026 21:32
Moving Claude Code work to Codex without rebuilding the harness

Moving Claude Code work to Codex without rebuilding the harness

The painful part of switching from Claude Code to Codex is not the model swap. It is the harness.

Claude Code stores useful operating context in ~/.claude: CLAUDE.md, skills, agents, hooks, MCP config, and sometimes session state. Codex has a different shape under ~/.codex. Copying files by hand works for the obvious pieces and misses the edge cases.

The first-party Codex import flow is the right starting point. The second pass is validating the pieces that an import checklist cannot prove:

  1. Hook behavior that does not port 1:1.
  2. Per-agent restrictions that Codex cannot enforce the same way.
@unitedideas
unitedideas / claude-to-cursor-migration.md
Last active April 18, 2026 20:28
How to migrate your Claude Code agent to Cursor (or any other AI tool) — manual recipe + file-by-file mapping

How to migrate your Claude Code agent to Cursor (or any other AI tool)

TL;DR — Every AI coding tool stores your agent differently. Claude Code uses ~/.claude/ with CLAUDE.md + agents/ + skills/. Cursor uses .cursor/rules/*.mdc. Codex uses ~/.codex/. Moving between them by hand means rewriting your system prompt in a new format and hoping you didn't miss anything.

Below is the manual recipe for the most common migration — Claude Code → Cursor — including every file that needs to move, the format differences to watch out for, and the subtle things that will silently break if you copy-paste. At the end I link to a one-command tool that does this automatically, because after doing it by hand twice I stopped doing it by hand.

The file layouts

@unitedideas
unitedideas / agent-ready-sites-wave.md
Last active April 18, 2026 03:10
120+ agent-ready sites added to Not Human Search in Q2 2026 — categorized, linkable, searchable via MCP

160 agent-ready sites — Q2 2026 curation wave

The Not Human Search index added 160 new agent-ready sites between 2026-04-17 and 2026-04-18 (9 submission waves of 20, rate-limited ~20/hr per IP). All are searchable + scored 0–100 on agentic readiness signals (llms.txt, OpenAPI, MCP, ai-plugin.json, structured API, robots.txt AI rules, Schema.org).

# Wire NHS into your agent (Claude Code, Cursor, Cline, Continue)
curl -fsSL https://nothumansearch.ai/install | sh

Then search live from your agent: "find MCP servers for vector databases", "agentic APIs for payments", "AI eval tools", "agent memory platforms".

@unitedideas
unitedideas / mcp-verify-oneliners.md
Created April 18, 2026 01:38
Verify any MCP server in 3 curls — Q2 2026 ecosystem health data, one-line probes

Verify any MCP server in 3 curls

If you build agents, you've probably hit this: a site advertises MCP support, but when you actually wire it into Claude Code, Cursor, or your own client, the /mcp endpoint either doesn't exist or doesn't answer JSON-RPC correctly.

Q2 2026 data: of 5,578 sites advertising MCP support, only 10.3% (575) actually respond to a live tools/list request. Text-mention claims outnumber working endpoints roughly 4:1. Source: 8bitconcepts MCP ecosystem health report.

Here are three one-liners to verify any MCP server before integrating.

1. Direct JSON-RPC probe (the ground truth)

@unitedideas
unitedideas / ai-hiring-geography.csv
Last active July 6, 2026 16:36
AI Hiring Geography — Top cities + salary by location, Q2 2026
city region role_count share_pct companies avg_salary_usd salary_n
San Francisco Bay Area US-West 2011 47.22 221 256967 1487
New York US-East 510 11.97 110 227606 387
Remote (US) Remote 309 7.26 84 219597 270
London Europe 205 4.81 56 258488 18
Remote (Global) Remote 180 4.23 59 202062 111
Boston US-East 159 3.73 21 200084 142
Seattle US-West 124 2.91 33 201321 105
Bangalore APAC 122 2.86 35 0 0
Austin US-South 93 2.18 16 212853 66
@unitedideas
unitedideas / ai-entry-level.csv
Last active July 6, 2026 16:15
The Junior AI Hiring Gap — Entry-level vs senior roles, Q2 2026
experience_level role_count share_pct
junior 551 6.28
mid 2169 24.71
senior 3025 34.47
lead 2429 27.67
principal 603 6.87
@unitedideas
unitedideas / ai-workplace.csv
Last active July 6, 2026 16:00
AI Engineering Workplace Data — remote vs onsite vs hybrid salary and role distribution
workplace_type role_count share_pct avg_salary
onsite 4704 53.59 221246
remote 2530 28.83 224637
hybrid 1543 17.58 263481
@unitedideas
unitedideas / state-of-ai-engineering.md
Last active July 6, 2026 15:45
State of AI Engineering Q2 2026 — Live snapshot of hiring, compensation, and infrastructure verification. Combines aidevboard.com + nothumansearch.ai data

State of AI Engineering — Q2 2026 Live Snapshot

Updated: 2026-07-06

A single reference combining hiring, compensation, and infrastructure-verification data from the AI engineering ecosystem. Sources are live public APIs from aidevboard.com and nothumansearch.ai.

Topline numbers

Metric Value
@unitedideas
unitedideas / ai-compensation-by-skill.csv
Last active July 6, 2026 15:30
AI Engineering Compensation by Skill — Live data from aidevboard.com/api/v1/stats
rank tag count avg_salary
1 research 738 278828
2 search 645 275537
3 reinforcement-learning 542 272576
4 distributed-systems 1449 261081
5 infrastructure 612 255571
6 pytorch 994 253033
7 fine-tuning 803 252091
8 llm 2590 246170
9 machine-learning 586 244423
@unitedideas
unitedideas / mcp-ecosystem-health.csv
Last active July 6, 2026 15:15
MCP Ecosystem Health Dataset — 10.3% of claimed MCP servers survive live JSON-RPC probe. Q2 2026 snapshot from nothumansearch.ai
rank name category agentic_score url mcp_verified
1 Fia Signals x402 finance 90 https://nothumansearch.ai/site/x402.fiasignals.com true
2 Lodi Kids Activities security 55 https://nothumansearch.ai/site/lodikidsactivities.com true