Associative Trails / writing
Writing
How professional firms get real value from AI without losing what they know: context it can use, answers that don't just agree, and work you can verify.
Why hard things look easy
TikTok removes 178 million videos a quarter. It still cannot reliably tell you whether a car is driving on the wrong side of the road.
Your second brain will not scale
A folder of markdown files will run a one-person business. It will not run a practice of fifty.
Where does the load go?
Your practice can keep every graduate post it has and still lose the training pipeline.
Cast in place
Your instructions, your model and your file format change at different speeds. One interface locks them together.
Everyone adopted, nobody designed it
AI reached almost every firm and changed almost nothing, because it arrived one person at a time.
Borrowing against the day a client asks why
Polished AI output no longer means good thinking. The gap between what your team ships and what it understands is quietly compounding into debt.
- 30Model churn is not your problemAI · 16 Jul 2026
- 29Introducing Ookbook - data that knows how to fetch itselfAI · 9 Jul 2026
- 28The canary in your context windowAI · 2 Jul 2026
- 27The Deliverable FallacyConsultancy · 25 Jun 2026
- 26I thought the post was finished. My readers disagreed.AI · 18 Jun 2026
- 25The Effort DividendAI · 11 Jun 2026
- 24Introducing Roundtable - A panel of experts that never clocks offAI · 4 Jun 2026
- 23What to Write Down (And What to Protect)AI · 28 May 2026
- 22What your AI tools are missingAI · 21 May 2026
- 21The yes-man in the room - how to design sycophancy out of your AI workflowAI · 14 May 2026
- 20Agents cannot do osmosisAI · 7 May 2026
- 19Tacit debt and the verification gapAI · 30 Apr 2026
- 18The next bottleneckAI · 23 Apr 2026
- 17Something Worth ManipulatingAI · 16 Apr 2026
- 16The agreement incentiveAI · 9 Apr 2026
- 15Human in the loopAI · 2 Apr 2026
- 14AI agent skills as pluggable knowledgeAI · 26 Mar 2026
- 13Experiments in JupyterPractice · 27 Oct 2023
- 12Streamlining credit checking: How Sling can helpPractice · 19 Apr 2023
- 11Azure functions help carry the loadPractice · 8 Mar 2023
- 10Identify users the Azure wayPractice · 2 Feb 2023
- 09Migrating a Coldfusion application to Microsoft Azure - setting up a data sourcePractice · 29 Sep 2022
- 08Classifying content in bulkPractice · 16 Sep 2022
- 07Dealing with short signoff delays in SlingPractice · 8 Sep 2022
- 06Responding to prospective client emails promptly and with confidencePractice · 20 Jun 2022
- 05Deep Learning Architect - Classification for Architectural Design through the Eye of Artificial Intelligence (2018)Practice · 2 May 2022
- 04Architectural Drawings Recognition and Generation through Machine Learning (2018)Practice · 1 May 2022
- 03OPTIMISING IMAGE CLASSIFICATION - Implementation of Convolutional Neural Network Algorithms to Distinguish Between Plans and Sections within the Architectural, Engineering and Construction (AEC) Industry (2019)Practice · 30 Apr 2022
- 02Drawing Recognition. Integrating Machine Learning Systems into Architectural Design Workflows (2020)Practice · 29 Apr 2022
- 01Classification of Building Information Model (BIM) Structures with Deep Learning (2018)Practice · 28 Apr 2022