# Write the Docs Berlin 2026 MOC > Event map. Topic home: [[3. Resources/MOCs/Documentation & Technical Writing MOC]] · Index: [[3. Resources/MOCs/MOCs|Maps of Content → Events & Conferences]] A map of the Write the Docs Berlin 2026 talk recordings captured to Reader and Obsidian on 2026-09-23: 13 full talks and 2 lightning-talk compilations (13 short talks), all published by Write the Docs on 2026-09-22. This is a **session index + thematic map**: where each talk sits, what it argued, what it measured, and which durable claims it fed. ## Core Question What did Write the Docs Berlin 2026 say about how documentation work survives two pressures at once, AI tools that make drafting cheap and headcount that stays flat, and which of the speakers' methods came with evidence rather than anecdote? > [!note] Each linked note carries the video embed, the 20-section deep-read notes, and the auto-caption transcript. Captions garble most proper nouns; each note lists its corrections. Timestamps below are transcript positions. --- ## 📊 Proving value and choosing what to measure The metrics cluster. Three talks agree that a docs number is worthless until it has a job, and differ on which job comes first. - [[3. Resources/Readwise/Articles/Proving and Improving the Value of Your Docs - Tal Gluck]] — the State of Docs research (1,100 respondents, 30 interviews; 1 in 4 teams track nothing) sorted into two metric jobs, proving value upward and improving the docs; his three-step start for teams tracking nothing (00:23:47), and "15 minutes of conversation beats an afternoon of analytics" (00:25:14) - [[3. Resources/Readwise/Articles/Doc Health Metrics How Two Writers Built a Framework to Prioritize Docs - Anastasia Suboch]] — a five-metric health framework (coverage by Diátaxis type, freshness as doc updates over code merges, discoverability by FAQ retrieval, plus two more) with named bands and a business consequence per band; AI-instrumented at about $50 per assessment, 3,500 lines of vibe-coded Python, presented as green/yellow/red to product owners - [[3. Resources/Readwise/Articles/Silence Is Not Approval When Better Docs Create More Feedback - Uladzislava Lahun]] — every new feedback instrument raised complaint volume while ratings stayed flat (00:15:29); a quiet channel measures distance from the docs, not satisfaction with them (00:18:28). The ratings drop that is the talk's empirical core lives on a slide the transcript cannot see ## 🏗️ Capacity, infrastructure, and repository architecture What to do when product complexity grows faster than the docs team. - [[3. Resources/Readwise/Articles/Good Enough at Scale Documentation Infrastructure Under Capacity Constraints - Elena Baskakova]] — two writers to nine products; four prioritization tiers, a six-field gap register, and the rule that every gap gets a decision even when it gets no fix (00:17:07); infrastructure over writing speed because writing "was never the real bottleneck" (00:05:12) - [[3. Resources/Readwise/Articles/Mapping Your Docs-as-Code Repo Navigating Architectural Evolution of Repositories - Hridyesh Bisht]] — three layouts (docs beside code, per-product split, central docs repo) weighed on ownership, access boundaries, localisation, and duplication; "prioritize your organization needs and boundaries first. The architecture follows" (00:15:09); the 30-branch versioning failure mode (00:19:32) - [[3. Resources/Readwise/Articles/Structure After the Fact Bringing Content Strategy to Existing Docs for a Design System - Jen Buley]] — first content person inheriting 700 pages from about 200 contributors; four costed research inputs, the hedging-language audit (00:16:07), and the finding that guidance was adopted only once it lived in the contributors' own toolchain (00:13:10); traffic up about 35% after the restructure, attribution flagged as post hoc ## 🤖 AI and the writer's judgment Three talks on where the human sits once generation is cheap. None offers outcome data; the operational content is what transfers. - [[3. Resources/Readwise/Articles/Writing Less, Owning More The Shift in Technical Writing - Whitney Burkes]] — prompts as documentation assets: curated, versioned, reviewed in pull requests, and tested across three or four repositories before they count (00:21:11); repository-agnostic wording (00:22:36); "own the input, own the output" (00:19:05). Layoff figures and a Microsoft study are cited without sources - [[3. Resources/Readwise/Articles/Beyond Automation Human Judgment as the Strategic Core of Business Writing - Kel Raleigh]] — the Judgment Over Generation model: three lanes (AI generation, human-AI interpretation, human judgment), six pre-prompt questions, five editing passes; the central behavioral claim, that review compresses as output gets more fluent (00:08:53), is asserted without citation - [[3. Resources/Readwise/Articles/Write Inclusive, People-First Language About AI - Alexandra Klepper]] — against "accessibility for AI": accessibility exists for human beings (00:19:48); optimize docs for the human users of agents, not for agents (00:29:14); state guidance for probabilistic tools without guaranteeing it (00:28:29). Five statistics arrive unsourced; the talk itself shipped with auto-captions and no chapters ## 🛡️ Trust, security, and the API contract Documentation as infrastructure that other people depend on, and as a surface that can be attacked. - [[3. Resources/Readwise/Articles/Trading in Trust How Documentation Protects People, Not Just Projects - Stephan Rayner]] — two workplace stories from opposite sides of the same dynamic; "documentation externalizes authority. It levels the playing field otherwise decided by power, clout, seniority, personality" (00:12:27); a four-pillar framework with a test question per pillar. Story two shows a rule was enforceable, not that it was correct, which he concedes at 00:09:35 - [[3. Resources/Readwise/Articles/When Words Become Attacks Prompt Injection Through a Technical Writing Lens - Oishi Banerjee]] — documentation, error messages, and support text flow into agents, so they are an attack surface; "the assistant is not broken but the channel is" (00:03:44); "writing a system prompt is technical writing" (00:12:36). The three attack examples are constructed, and the 66.3% human-review figure sits on a slide with no spoken source - [[3. Resources/Readwise/Articles/Documenting APIs New Era, Same Craft - Javier Rubio]] — REST has no contract, so "the contract are the docs" (00:03:03); OpenAPI, its extensions, and the Arazzo workflow spec; specs generated from tests so drift goes red in CI (00:14:50); a relayed claim that more than 66% of one vendor's docs traffic in July was non-human (00:25:14), recorded as unsourced ## 🚀 Career and craft - [[3. Resources/Readwise/Articles/Moonshot Mindset A Tech Writer's Guide to Thinking Big and Breaking Free From the GDZ - Kate Wilcox]] — the "good-dog zone" of reactive request-filling versus strategic partnership (00:06:34); the CRAFT framework (Concept, Roadmap, Alignment, Flight, Telemetry) and a 10%-of-the-week budget in protected blocks; two case studies with outcomes stated as adjectives, no numbers ## ⚡ Lightning talks Two compilations, thirteen talks. The strongest transferable claims are pulled out; the rest are indexed in each note. - [[3. Resources/Readwise/Articles/Monday Lightning Talks - Write the Docs Berlin 2026]] — Deepak Valagam: "the machine is not the skill... the judgment is the skill" (00:09:42); Emily Axel on writing the longer invitation because those who don't need it skip it (00:14:02); Moronfolu Olufunke on owning your data versus hosting it (00:23:26); Fraser Keddie on editing as the invisible art (00:29:52); plus Florian Stolzenhain and Jonathan Putnam - [[3. Resources/Readwise/Articles/Tuesday Lightning Talks - Write the Docs Berlin 2026]] — Brooke Stauner: the reader "is doing so because they're having a really really bad day" (00:03:39); Gugulethu H. on deciding whether to call a model at all before routing it (00:18:50); Erica Beyea on the 25-hour week as a fair wage, not a perk (00:24:39); Anton Kolesnyk, Kenneth Whyte, Miranda Wilson, Val Grimm --- ## 🧠 Permanent notes this conference fed - [[3. Resources/Permanent Notes/A Docs Metric Earns Its Place by Naming Which Job It Does]] — from Gluck - [[3. Resources/Permanent Notes/A Documentation Gap Is Acceptable Only Once It Is a Recorded Decision]] — from Baskakova - [[3. Resources/Permanent Notes/Documentation Prompts Are Documentation Assets Governed Like Code]] — from Burkes - [[3. Resources/Permanent Notes/Any Text an Agent Reads Is a Control Surface]] — from Banerjee - [[3. Resources/Permanent Notes/Absence of Complaint Is Resignation Not Satisfaction]] — Lahun's talk became new evidence for this existing 2026-08-24 claim rather than a duplicate Candidates not yet written (in each note's Refactor Appendix): Rayner's "write policy for the person who must enforce it," Buley's "guidance lives in the contributors' toolchain," Rubio's "a spec that can't fail CI is a hand-written spec," Suboch's "a dashboard metric needs a business consequence per band," Wilcox's "manager approval is a gate, not alignment," Raleigh's "ask the judgment questions before the prompt." ## 🔗 Related maps - [[3. Resources/MOCs/Documentation & Technical Writing MOC]] — the topic map this event hangs off; the claims that outlive the conference are registered there - [[3. Resources/MOCs/DevRel MOC]] — the docs-metrics and gap-register claims are registered there; developer documentation is DevRel's largest surface - [[3. Resources/MOCs/AI Agent Safety MOC]] — Banerjee's control-surface claim and the stopping-point rule sit there - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — Burkes and Raleigh on where the human sits once generation is cheap - [[3. Resources/MOCs/Accessibility & Disability Inclusion MOC]] — Klepper's argument about who accessibility is for - [[3. Resources/MOCs/Evidence & Claims Literacy MOC]] — the recurring pattern across the series: precise-sounding figures (66.3%, 66%, 35%, 1 in 4) arriving without method or sample ## Open threads - The two lightning-talk notes and the Rayner note name speakers whose captions could not be resolved; verify against the conference schedule before citing anyone by name. - Session pages and slide decks were not captured; every claim that "lives on a slide" above is unverified from the transcript alone.