The stack I actually work in
Tools, models, and surfaces I use every day. Not aspirational — only things I've shipped real work with. Updated when the workflow changes, not on a calendar.
Daily drivers
What I open every morning
Area 1
AI coding agents
Where most of the building actually happens. Day-in, day-out drivers for shipping production code.
- Daily driver
Primary driver. Most production builds, refactors, and long-running agentic work go through here.
- Daily driver
Second seat for parallel work — different model, different context, useful for cross-checking and longer-horizon tasks.
- Regular use
Quick edits in an IDE that knows my codebase. Favored for small, surgical changes.
Area 2
Conversational models
The thinking surface. For drafting, scoping, brainstorming, and rubber-ducking before a build starts.
- Daily driver
Default for thinking. Long-context Projects with my context library loaded for ops, marketing, research, finance, and engineering threads.
- Daily driver
Second opinion. Different model, different bias profile. Useful for cross-checking before committing to a direction.
- Daily driver
Long-context grunt work. Drop in 200 pages of source material and ask the obvious question — Gemini handles it without flinching.
Area 3
Research & playgrounds
Where I learn, fact-check, prototype, and benchmark — before anything ships into a build.
- Daily driver
First stop for almost any research question. Cited answers, follow-up threading, and a real link list at the bottom — way faster than search.
- Daily driver
Where source-grounded research happens. Drop in PDFs, transcripts, and threads, then ask questions against just that corpus. The audio-overview feature is genuinely useful for long material.
- Daily driver
Gemini prototyping. Where I test long-context patterns and structured-output schemas before moving to API. Free, fast, surprisingly capable.
- Built deeply with
Prompt design, evals, and Workbench experiments before anything ships into a Claude-driven app.
- Regular use
Quick model-vs-model checks and tool-use prototypes against the OpenAI API.
Area 4
Daily surfaces
Where the day's coordination, writing, and dashboards live — alongside the editor and terminal.
- Daily driver
Docs, Sheets, Calendar, Mail. Where the writing, planning, and coordination happens when it isn't in code.
- Daily driver
A handful of internal tools and dashboards I've built for the daily workflow — home telemetry, personal metrics, custom indexes. None of them ship to anyone but me.
- Daily driver
Default editor. Tuned with the agent integrations I rely on day-to-day.
- Terminal (iTerm2 + tmux)Daily driver
Where Claude Code lives. Long-running sessions across multiple panes.
- Regular use
When the build is iOS-side (Bean Dialer, Twin Talk shell prototypes).
Area 5
Data analysis
Where I dig into numbers — pulling, transforming, modeling, and visualizing data to answer real questions instead of guessing.
- Regular use
Default for any non-trivial data question. BigQuery for warehouse-scale, Postgres when the data already lives there.
- Regular use
Where the analysis happens once SQL has done the heavy lifting. Notebooks for exploration, scripts for repeatable pipelines.
- Regular use
Exploratory analysis surface. Where I rough out a model, sanity-check a hypothesis, or build a one-off visualization.
- Google Sheets (advanced)Regular use
Underrated. For one-off models, quick collaborations, and anything where a non-technical stakeholder needs to read and edit. Pivot tables, ARRAYFORMULA, QUERY().
Area 6
Marketing tools
The day-to-day SEO, search, and analytics tools any marketer with a technical practice ends up living in.
- Regular use
Default keyword and competitive-research surface. Position tracking, content gap, the whole pipeline.
- Regular use
Second seat for backlink data and competitive intelligence. Different crawler, different index, useful for cross-checking Semrush.
- Daily driver
Source of truth for how Google actually sees a site. Performance, indexation, Core Web Vitals, structured-data validation.
- Regular use
Behavior, conversion, and attribution data when the site has it instrumented properly.
- Regular use
Technical SEO crawler. The first tool I reach for on any new site audit — indexation, redirects, canonicals, schema, rendered vs raw HTML.
- Regular use
Free visualization layer over GA4, Search Console, BigQuery, Sheets. Dashboards for stakeholders who don't want to live in the source tools.
Area 7
Automation & glue
What I reach for when the answer is 'wire two systems together' instead of 'build a new one.'
- Built deeply with
Self-hosted workflow automation. Long-running pipelines that need to be inspected, paused, and reasoned about.
- Regular use
Boring no-code glue between SaaS tools. Right when the workflow is small and the integrations are official.
- Airtable + Notion APIsRegular use
Whenever the data needs to live somewhere a non-engineer can read and edit.
Area 8
Infra & deploy
Where the artifacts of the work get stored, deployed, and observed.
- Daily driver
Deploy target for everything web. Edge functions for the proxy/CDN-layer logic.
- Regular use
Postgres + auth + storage when the build needs persistence behind a thin API.
- Pinecone / WeaviateRegular use
Vector stores when retrieval is core to the product. Pinecone for managed, Weaviate for self-hosted.
- Regular use
Time-series storage for the home-automation telemetry stream and anything else where the derivative matters more than the value.
Reading this list
What the tags mean
Each tool has a proficiency tag. Here's what each one actually means.
Open it most days. If it broke for a week, my workflow would too.
Shipped non-trivial production work with it. Know its sharp edges.
In rotation but not every day. Reach for it when the job calls for it.
If a tool isn't on this page, it's either off the island or hasn't earned a seat yet. The list evolves — additions are intentional, demotions come off the same day.