Tools I use in practice.

My thinking layer for product reasoning, PRDs, strategy, and rapid research, and for comparing different perspectives on complex questions.

Cheaper open-weights models I reach for as a second opinion and when it comes to speed and low to medium-scale tasks (recently on various side projects).

Where I source open models and datasets, and watch what the community ships before it reaches the major APIs.

Used to interrogate source-heavy material and synthesize reports and papers against a controlled document set.

My system for requirements, roadmaps, documentation, project tracking, and keeping execution visible.

Cowork agents with custom skills for automating workflows and tasks across documents, analysis, planning, and everyday knowledge work.

My vibe-coding and agentic development stack for planning, implementation, larger changes, code review, and multi-step execution.

The IDE I use daily, paired with Copilot as an agentic code companion for refactors, debugging, and staying close to the code while iterating.

The local runtime I use when a prototype needs to stay private or stay cheap.

The core stack for cleaning data, exploring relationships, automating analysis, and building evidence-backed models.

My modeling toolkit for interpretable machine learning, anomaly detection, evaluation, and neural experiments when deeper architectures add value.

For products that need retrieval, tool use, domain context, or multi-step reasoning instead of a single model response.

I turn models into usable interfaces early, so stakeholders can react to behavior rather than abstract specifications.

Where data products, dashboards, and visualization get turned into decisions, showing where the numbers disagree, where the risk sits, and which trade-off we're making.