The handbook thesis, defended.
Why we ship a long-form document at the end of every engagement, and why the slide deck is about to die as a consulting artifact. Featured.
W6 Archive
The archive. Every essay we have shipped.
Twenty-four essays, four pillars, one voice. Filter by pillar or scan by date. Essays we have not yet published appear with a muted byline and a draft indicator. We link only the ones that are shipped.
Prompt caching changed our unit economics.
When a 140 page handbook is generated against the same 20k token context forty times, the cache pays for the whole engagement.
Why we stopped writing one-shot prompts.
Every deliverable runs through four stages now: draft, self-critique, human edit, eval.
Extended thinking is a research budget, not a toggle.
When to turn on extended thinking, when to pay for it in tokens, and how to know if your prompt actually needs it.
How we write templates that survive a model upgrade.
Our template library has been through three Claude versions without a rewrite. The principles that make a template portable.
What tool use unlocks that chat-only Claude cannot do.
Concrete consulting examples: a retrieval-grounded pentest reviewer, a code-reading auditor, and a topical map generator.
The self-critique loop: how we get Claude to find its own mistakes.
Prompt patterns and real traces from our handbook production pipeline.
The 200k window is a context engineering problem, not a memory problem.
Retrieval patterns for long-running engagements. Why bigger contexts do not remove the need for careful assembly.
The pentest report is a literary form, and most of them are bad.
Sixty percent of pentest reports we inherit are unreadable to the executive and unactionable to the engineer.
The IAM attack path every Series A SaaS eventually has.
One policy, one role, one bucket. The same misconfiguration shows up at seventy percent of Series A SaaS companies we audit.
Claude is not a pentester. It is a very fast junior analyst.
How we use it without letting it run the engagement.
Threat modeling goes from 3 days to 3 hours when Claude drives STRIDE.
A workshop trace with actual prompts, actual diagrams, and actual decisions.
The incident response runbook nobody wrote because nobody had time.
Template plus how Claude shortens the write-up without hiding the decisions that matter.
OWASP LLM Top 10 translated into concrete audit checks.
A checklist for security readers who want to audit their own LLM-backed products without becoming AI researchers.
We measured share of answer for 400 queries across 4 LLMs.
A structural audit of what actually gets cited in Claude, ChatGPT, Perplexity, and Gemini answers. Data, not theory.
Citations are the new backlinks.
A tactical guide to being the source that LLMs quote, not just the page they read.
Why we stopped chasing Google featured snippets.
With ChatGPT and Perplexity absorbing the top-of-funnel, the snippet game is a loser's game. What we are doing instead.
How to build a topical map that ranks in Google AND gets cited by ChatGPT.
Dual optimization with one content model. Hub, spoke, and the citation surface in between.
Every crawler is an LLM now: the new robots.txt discipline.
Practical robots.txt and llms.txt patterns, with the list of crawlers we actually allow.
Billing hourly in 2026 is a moral hazard.
When an operator plus Claude can do a week of classical consulting work in a day, the hourly rate incentivizes slower delivery.
A consulting firm's P&L when the model is a team member.
Our actual cost structure in the first year, broken down line by line. Published so the next founder does not have to guess.
How to price a fixed-scope engagement when the work is 10x faster.
The pricing framework we use, including the two places we almost always underprice.
The operator-to-senior ratio at an AI-native firm.
Why we chose not to hire associates. What a senior plus a model produces that an associate cannot.
When to sell a retainer and when to sell a sprint.
The decision tree we actually use with prospects, with the three questions that settle it in under ten minutes.
We publish one essay a week on Tuesdays. Subscribe here to get each piece in your inbox.