Consulting and Craft: The Next Shelf

Outlined candidates for the craft half of the Thursday stream, written in the same spirit as NEXT-UTH-IDEAS.md and NEXT-STORY-IDEAS.md: to mull over, not to commit to.

This file exists because the Thursday slot is described as mixed between four series and is not. Measured October 2026:

for f in _posts/*.md; do grep -m1 '^series:' "$f"; done | sort | uniq -c

So the Thursday stream is single-series for its last eighteen months. Three of the four series run dry between April 2028 and March 2029 while Under the Hood keeps going, which is a shelf problem rather than an editorial decision: nobody chose to stop writing craft posts.

What the existing 50 cover: consulting practice (scoping, saying no, contracts, the discovery ledger, retrospectives, data-informed decisions, the planning onion, async and written-first), the kitchen-metaphor craft essays (Through the Kitchen), practical AI delivery (the Bedrock pilot run), and hands-on infrastructure recipes (a two-tier offline CA, email capture, HTTPS on localhost, ephemeral environments, a pocket identity provider, docs that fail the build). The gaps below were chosen against that list.

Reading Code You Didn’t Write

The practice half of what The Lap dramatises. The narrative has the archaeology (The Dig Site, and the characterisation tests that follow it); the craft series has nothing a reader can apply on Monday to a codebase they have just been handed. Sections: what to establish in the first day and in what order (how it builds, how it deploys, how it is tested, what the money path is); tracing one real request end to end as the cheapest map available; reading the commit history as a document (who left, what got reverted, where the churn is); finding the seams before changing anything; characterisation tests as a way to write down what the system does rather than what it should do; the dead-code question and how to answer it with evidence instead of courage; what to write down for the next person, and when. Links back to “The First Hour of Every Project”, “Code Is Read More Than It Is Written” and “Technical Debt Is a Loan, Not a Crime”. The single largest hole: half the blog’s narrative is about inherited systems and the craft series only ever starts greenfield.

The Restore You Haven’t Tried

Hands On. Backups are the one piece of infrastructure practice the series never covers, and the failure is always the same: the backup ran, and nobody has ever restored it. Sections: what a backup is for, stated as recovery objectives before any tool is chosen; the three failures that actually happen (the backup covered the wrong thing, the restore needs a credential that is inside the thing being restored, and the restore works but takes four days); a restore drill as a scheduled exercise with a written outcome; restoring to a different place than production, so the drill is not itself an outage; what to test beyond the data (schema, secrets, DNS, the runbook); ransomware as the case that makes immutability and an offline copy the design rather than the upsell; the drill cadence a small team can actually keep. Links back to “Ephemeral Environments on a Budget” and “Documentation as a Byproduct”.

Test Data That Isn’t Production

Hands On, and a problem nearly every team solves badly once and never revisits. Sections: why a copy of production is the easy answer and what it costs the moment anyone outside the team needs an environment; the three approaches, with the trade-off stated plainly (anonymisation, synthesis, and a small hand-built fixture set); referential integrity as the thing that breaks every naive anonymiser; the fields that re-identify people even when the obvious ones are gone; why synthetic data has to inherit the shape of the real distribution or the tests pass on data nobody could have; keeping a fixture set alive as the schema moves; what to do about the one bug that only reproduces on real data. Links back to “Spikes That Can’t Leak”, “Ephemeral Environments on a Budget” and the Under the Hood post on anonymity.

When They Want a Date

Consulting practice, and the conversation the series currently steps around. “How Estimation Works (And Why It Doesn’t)” explains why estimates are wrong; “Scoping the Work” and “The Planning Onion” cover the shaping; nothing covers the moment someone asks for a date and will not accept a range. Sections: what the question is usually really asking (a decision date, a budget, a committed external promise, or reassurance), and how to find out which; why a single number is the one answer that carries no information, and what to offer instead; forecasting from throughput rather than from estimates, worked on a small real example; the difference between a date you can commit to and a date you can forecast, and how to say which one you are giving; what to do when the date is fixed first and the scope has to move; how to report a slip early, in the terms the person asking actually cares about. Links back to “How Estimation Works (And Why It Doesn’t)”, “The Planning Onion”, “Scoping the Work”, “Directional, Not Definitive” and “Contracts Are a Conversation You Have Early”.

Leaving Well

The end of an engagement, which the narrative covers (The Handover) and the craft series does not. Sections: designing for your own removal from the first week, and why that is a deliverable rather than a sentiment; what actually has to transfer (the decisions and their reasons, not the code, which they already have); the handover artefacts worth writing and the ones nobody reads; running the last month as a drill where the team does the work and you watch; the awkward commercial truth that a good exit shortens the engagement, and how to be straight about it; what to leave behind when the client cannot hire a replacement; staying available without staying load-bearing. Links back to “Documentation as a Byproduct”, “Scaling Knowledge”, “Contracts Are a Conversation You Have Early” and “Saying No”.

Kept warm

• The Incident You Are Not Fixing. Being on call, the comms during, and writing a timeline that survives a reader who was not there. The Workshop covers blameless post-mortems; the hours before one are uncovered.
• Observability You Can Afford. The craft companion to the 2030 Under the Hood post: what to instrument first on a small budget, and why cardinality is the bill. Schedule after it.
• Secrets for Small Teams. Hands On. Sits naturally with the CA run and the pocket identity provider.
• The Review That Is Worth Having. Code review as a practice rather than a gate, and what to do about the reviewer who is also the bottleneck.
• High Performance Teams needs a shelf of its own. Five posts, last one April 2028, and the series has a defining post, a measurement post and a psychological-safety post with nothing between them. Worth deciding whether it continues or folds into Consulting and Craft.