Review2026.05.24 · 11 min read
The workflows that stuck, the workflows that didn’t, and the question of what a primary thinking partner is for.
AI assistant
Fourteen months in. Claude Pro is the most load bearing piece of software in my stack — more than any productivity tool I’ve used at length, including the ones I built habits around for years before LLMs were credible.
Anthropic’s subscription tier to their Claude model family. Long context window, Projects feature for persistent reference material, Artifacts for code and structured output.
I run a morning workflow: Claude reads my calendar, my Linear queue, and the unread Slack threads I’ve starred, and produces a daily briefing. It writes first drafts of long emails I then edit. It runs research passes on topics I’m about to write about. It helps me debug code I’m not strong in. It’s the second brain I never managed to build in Obsidian alone.
Direct comparison vs GPT 4o and Gemini, run consistently over six months on the same prompts: Claude wins on long context reasoning, on writing that reads as written by a person rather than a model, and on refusal patterns that feel like editorial judgment rather than safety theatre. GPT 4o wins on speed and on multimodal. Gemini wins on integration with Google Workspace.
Anyone whose work involves writing, research, code, or long form thinking, at any scale where $20/month doesn’t register as a cost. The break even is roughly one hour per month.
People who want a chatbot. The value compounds when you treat it as a thinking partner with a system around it (Projects, persistent context, your own prompt library). If you’re only ever asking it one off questions, the free tier is fine.
Three Projects: one for the platform (Brand voice, pillar definitions, refusal list), one for research (with citations and pending questions), one for ops (calendar, inbox, action items). Each Project has 5–15 reference files. Same morning briefing prompt, refined monthly. Same writing prompt for first drafts.
Writing throughput up roughly 3x at equivalent quality. Email triage time down by 70%. Research depth on weekly pieces up substantially — I now actually read primary sources because Claude saves me the time I’d have spent paraphrasing them. What didn’t shift: the quality ceiling. Claude can’t make me a better writer; it can only remove the friction between me and the writing.
— Alvin
A note on this review. This entry sits inside the AI & Digital pillar of The Human Upgrade. It is an n=1 working log, not medical advice. Alvin Tan is a functional health coach in training, not a licensed clinician. The Human Upgrade may earn a commission on purchases made through the link above; disclosure does not change verdicts. Any reader considering interventions should consult a qualified clinician in their own jurisdiction.