THE MAY STACK

The AI tools we actually use.

Re-tested every month. Some are paid, some are free-tier, and all are here because they have a real job in the workflow. No sponsorships. If a tool no longer fits, we remove it.

AI Tool
What it's for
Verdict
Rating / Plan
Claude
Reasoning / Writing
Our default for long-context thinking, careful writing, PDF review, and turning messy notes into usable operating docs.
Best for · strategy, writing, analysisSkip if · fresh source research
●●●●●
$20/mo+
Wispr Flow
Voice / Capture
Voice input across the computer. We use it to think out loud, draft faster, and capture rough ideas before they get over-edited.
Best for · dictation, rough draftsSkip if · quiet shared spaces
●●●●●
Free / $12+
Cotypist
Autocomplete / Typing
Everywhere autocomplete on the Mac. We use it a lot because it follows typing across apps and the latest Qwen models make the suggestions feel useful.
Best for · typing everywhereSkip if · non-Mac workflows
●●●●●
Free beta
Krisp
Meetings / Audio
Noise cleanup and meeting audio support for calls where the recording, transcript, or client experience needs to be cleaner.
Best for · calls, noisy roomsSkip if · already clean audio
●●●●●
Trial / $8+
Notion
Workspace / Ops
The workspace for tasks, notes, client context, content calendars, and the durable operating layer around AI-generated work.
Best for · systems, documentationSkip if · fast one-off notes
●●●●●
Free / $10+
ChatGPT
General / Multimodal
The broad second tool in the stack: quick ideation, image work, voice, broad Q&A, and a second opinion against Claude outputs.
Best for · multimodal, quick passesSkip if · final factual sourcing
●●●●●
Free / $20+
Codex
Coding / Agents
OpenAI's coding agent as its own app and workspace. We use it for repo work, tests, implementation loops, and code changes that should not stay trapped in chat.
Best for · repo work, tests, PRsSkip if · non-code workflows
●●●●●
ChatGPT plan
Cloudflare Workers
Backend / Edge
Small, fast backend logic at the edge: APIs, cron jobs, webhooks, security gates, and glue code that should not need a full server.
Best for · edge APIs, automationSkip if · stateful app servers
●●●●●
Free / $5+
Vercel
Deployments / Web
Preview deployments and production hosting for modern web projects, especially when a Next.js site needs fast iteration.
Best for · frontend deploysSkip if · non-web infrastructure
●●●●●
Free / $20+
Manus
Agents / Research
A general agent workspace for delegated research, browser-style tasks, slides, sites, and experiments that need more autonomy.
Best for · delegated tasksSkip if · sensitive accounts
●●●●●
Free / $20+
Cursor
Code Editor
Installed and useful, but not our paid coding environment. The free tier is enough because we do not do our main coding there.
Best for · light code inspectionSkip if · our primary coding flow
●●●●●
Free tier
Perplexity
Research / Sources
Source-backed research for current facts, market context, vendor checks, and the evidence pass before Claude does synthesis.
Best for · cited researchSkip if · private work data
●●●●●
Free / $17+
Google AI Mode
Search / Answers
Google Search's AI answer mode for quick, web-grounded answers with helpful links. We use it alongside Perplexity when accuracy and current information matter.
Best for · quick cited answersSkip if · long research threads
●●●●●
Free
Last re-tested May 1, 2026. We update this page on the first weekday of every month.