# The Token Department sitemap

Live pages and machine-readable files. Canonical HTML is for readers. Markdown twins contain the source text.

## Home & Overview

- [Home](https://thetokendepartment.com/) · The four inputs that shape useful work from AI.
- [Essays](https://thetokendepartment.com/essays/) · Published essays and forthcoming work.
- [Guides](https://thetokendepartment.com/guides/) · [Markdown](https://thetokendepartment.com/guides.md) · Practical guides to working with AI. Start with a workflow, build the system around the model, or choose models and capacity.
- [Running a Token Department](https://thetokendepartment.com/guides/running-token-department/) · [Markdown](https://thetokendepartment.com/guides/running-token-department.md) · Organise seven practical tools around model selection, routing, the harness and user skill.
- [The Token Department](https://thetokendepartment.com/essays/the-token-department/) · [Markdown](https://thetokendepartment.com/essays/the-token-department.md) · Manage model access, task routing, the harness and user skill as one continuing responsibility.
- [About](https://thetokendepartment.com/about/) · [Markdown](https://thetokendepartment.com/about.md) · A place to choose the models and token plans that fit your tasks and budget.

## Model Selection

- [Model selector](https://thetokendepartment.com/resources/ai-model-selector/) · An interactive comparison of models, token plans and plan value.
- [Model Selection](https://thetokendepartment.com/essays/model-selection/) · [Markdown](https://thetokendepartment.com/essays/model-selection.md) · Buy model access around forecast work, required quality and the capacity each plan can actually support.
- [Small models, running locally](https://thetokendepartment.com/guides/open-weight-models/) · [Markdown](https://thetokendepartment.com/guides/open-weight-models.md) · Choose a local model that fits your memory. Run private text tasks and free voice transcription on a laptop, Mac or desktop GPU.
- [Small language models](https://thetokendepartment.com/guides/small-language-models/) · [Markdown](https://thetokendepartment.com/guides/small-language-models.md) · Define a focused task, then test how much model it needs. Compare accepted output and total workflow cost on a stated configuration.

## Effective Routing

- [Model Routing](https://thetokendepartment.com/essays/model-routing/) · [Markdown](https://thetokendepartment.com/essays/model-routing.md) · Choose the exact model and effort that produce acceptable work at the lowest total cost.
- [Token Procurement](https://thetokendepartment.com/guides/token-procurement/) · [Markdown](https://thetokendepartment.com/guides/token-procurement.md) · Buy model access around your workload. Compare allocations, capacity, required quality and budget, with a complete worked example.

## Harness Quality

- [On pstack: the skills behind 2,500 PRs a month, and the mental models behind the skills](https://thetokendepartment.com/essays/on-pstack/) · [Markdown](https://thetokendepartment.com/essays/on-pstack.md) · Lauren Tan’s agentic engineering skills, and six pictures drawn from her essays and interview.
- [The Harness](https://thetokendepartment.com/guides/the-harness/) · [Markdown](https://thetokendepartment.com/guides/the-harness.md) · A harness is the software that manages context, tools and instructions around a model. Design it so the model is hot-swappable.
- [Interfaces](https://thetokendepartment.com/guides/interfaces/) · [Markdown](https://thetokendepartment.com/guides/interfaces.md) · Choose how to brief agents and review their work. Compare chat, workspace apps, terminals and Agent Development Environments (ADEs), with local and cloud options.
- [Building skills and workflows](https://thetokendepartment.com/guides/skills-and-workflows/) · [Markdown](https://thetokendepartment.com/guides/skills-and-workflows.md) · A skill stores a reusable method. A workflow connects the steps, evidence and checks needed to finish a piece of work.
- [Using GitHub](https://thetokendepartment.com/guides/using-github/) · [Markdown](https://thetokendepartment.com/guides/using-github.md) · Learn Git and GitHub through interactive diagrams of commits, branches, worktrees, shared reviews and releases.
- [Context managers and context engineering](https://thetokendepartment.com/guides/context-engineering/) · [Markdown](https://thetokendepartment.com/guides/context-engineering.md) · Context engineering decides what enters the model's window, in what order, and what stays on disk until the task needs it.
- [Connectors and tools](https://thetokendepartment.com/guides/connectors-and-tools/) · [Markdown](https://thetokendepartment.com/guides/connectors-and-tools.md) · The model requests an action. The harness checks permissions, runs the tool and returns the evidence needed for the next decision.
- [On-device and cloud agents](https://thetokendepartment.com/guides/on-device-and-cloud-agents/) · [Markdown](https://thetokendepartment.com/guides/on-device-and-cloud-agents.md) · Choose the interface, the environment that runs tools and the location of model inference separately. Each has its own access and data boundary.

## User Skill & Evals

- [Voice mode](https://thetokendepartment.com/guides/voice-mode/) · A guide to speaking instructions and drafting with AI.
- [Write your own test for AI](https://thetokendepartment.com/essays/write-your-own-test/) · [Markdown](https://thetokendepartment.com/essays/write-your-own-test.md) · [Alternate Markdown](https://thetokendepartment.com/essays/write-your-own-test-for-ai.md) · Running evals yourself, based on the way you work, to determine which models give you the best bang for buck.
- [Building AI Capability](https://thetokendepartment.com/essays/learning/) · [Markdown](https://thetokendepartment.com/essays/learning.md) · Develop people through bounded work, shared practice and responsibility for the result.

## Feeds

- [Jensen Says](https://thetokendepartment.com/jensen-says/) · Filterable quotes and exhibits on AI and computing.
- [AI Meta Meter](https://thetokendepartment.com/ai-meta-meter/) · An interactive view of AI arenas and signals.
- [Agentic Engineering on X](https://thetokendepartment.com/notable-on-x-coming-soon.html) · A research draft of builder posts on agent workflows.
- [AI in China](https://thetokendepartment.com/ai-in-china/) · [Markdown](https://thetokendepartment.com/ai-in-china.md) · Sources for the China research feed. The interactive HTML page carries the current feed records.

## Reference & Agent Endpoints

- [Skills](https://thetokendepartment.com/skills/) · [Markdown](https://thetokendepartment.com/skills.md) · Eight public agent skills and Modern Web Guidance for web platform research.
- [Glossary](https://thetokendepartment.com/glossary/) · [Markdown](https://thetokendepartment.com/glossary.md) · The recurring terms behind model selection, effective routing, harness quality and user skill.

- [Agent discovery](https://thetokendepartment.com/llms.txt) · A concise index for language models.
- [Markdown sitemap](https://thetokendepartment.com/sitemap.md) · The directory of live pages and agent files.
- [XML sitemap](https://thetokendepartment.com/sitemap.xml) · Canonical URLs for live pages.
- [RSS feed](https://thetokendepartment.com/rss.xml) · Published essays and guides in RSS 2.0 XML.
