AI & Agents · Strategy
The COO's Guide to Choosing the First Workflow to Automate
Five filters for choosing the first workflow to automate.
Read articleWriting
Notes from the foundry.
Field notes from two decades of building software and companies.
AI & Agents · Strategy
What CFOs Should Ask Before Approving Another AI Budget
CFOs should require five answers before approving AI budgets: measurable outcomes, accountable ownership, full operating costs, rollback, and mapped workflows.
Read articleAI & Agents
Using GPT-5.6 Luna at Max in Codex
Use Terra for coordinating Codex subagents, and Luna at XHigh or Max for bounded work in separate delegated threads that Sol reviews and combines centrally.
Read article
AI & Agents · Strategy
Your AI Strategy Is Not a Tool List
Build an AI strategy around agent-ready operations, not a tool list. Map workflows, constraints, permissions, ownership, and rollback paths before you buy.
Read article
AI & Agents
OCR is a routing problem now
OCR benchmarks are saturated and the leaderboards disagree. The winning 2026 architecture routes each document to the cheapest engine that can handle it.
Read articleAI & Agents
Codex CLI config.toml Guide 2026
Current Codex CLI config.toml guidance: safe defaults, profile files, sandbox and approval migration, AGENTS.md, subagents, plugins, and CI-ready settings.
Read article
AI & Agents
The self-driving company still needs people
A self-driving company is not a company without people. It is a company where people set direction and agents run accountable, verifiable operating loops.
Read article
AI & Agents
Building Private AI Evals: Why Your Company Needs Them and How to Start
No team can defend an AI workflow decision without measurement. A practical guide to private, versioned evals — with a manufacturing AR dispute example.
Read articleAI & Agents
The Codex operating manual for knowledge workers
Don't ask Codex to be strategic, give it an operating system. Eight rules for running knowledge work on durable threads, briefs, and human approval gates.
Read articleAI & Agents
Automations turn follow-up into managed background work
Knowledge work isn't hard, it's relentless. Codex automations turn recurring follow-up into background loops that keep work warm and surface only exceptions.
Read articleAI & Agents
Worktrees for knowledge work: parallel work without contamination
Developers use worktrees to run parallel work without contamination. Knowledge workers need the habit too: isolate each line of inquiry, then merge on purpose.
Read articleAI & Agents
The chief-of-staff thread: one persistent place to hold context
Skip the prompt library. Start with one persistent chief-of-staff thread that holds context and routes it into action, then have every project thread report in.
Read articleAI & Agents
Codex is becoming a workspace, not a chatbot
Codex isn't a better chatbot, it's a workspace. Persistent threads, background work, isolated worktrees, repeatable skills, and handoffs in one surface.
Read articleAI & Agents
Portable Agent Ops reduces coding harness switching cost
Stop storing trust in one coding harness. A portable Agent Ops layer (AGENTS.md, thin shims, shared skills, a certification suite) makes migration a wiring job.
Read articleAI & Agents
YAML is the workflow layer for AI agents
The next unlock in AI agents isn't a bigger model, it's a YAML file. Portable pack.yml skill bundles kill the setup tax and make named workflows run like tools.
Read article
AI & Agents
Cowork Is Not Chat. Stop Using It Like Chat.
Three products share the Claude name. Chat answers questions, Code ships software, Cowork completes tasks. Most people use a power tool as a screwdriver.
Read article
AI & Agents
AI as a Team, Not a Chatbot
One task, one agent, one result caps your throughput. The largest gains come from agents running in parallel with defined roles and peer-review gates.
Read article
AI & Agents
Custom Skills: Build Once, Run Forever
You explained your weekly report format to Claude. Then explained it again. And again. A five-section skill file ends the repetition permanently.
Read article
AI & Agents
The Monday Briefing That Runs Without You
Every Monday at 7 AM, Claude scans Slack and the calendar, flags what needs prep, and saves a briefing. Nobody typed a prompt. Nobody opened the app.
Read article
AI & Agents
Plugins Are Composable. Stack Them.
Most people install one Cowork plugin and stop. The real advantage comes from stacking two in a single task, where domain expertise meets data backing.
Read article
AI & Agents
The _MANIFEST.md: Five Minutes That Save Hours
Claude read your whole project folder, including the draft you scrapped three weeks ago. The output blends old thinking with new. The fix takes five minutes.
Read article
AI & Agents
Stop Telling AI What to Do. Tell It What Done Looks Like.
Most people write prompts the way bad managers give assignments. The best results come from describing the destination and letting the agent find the route.
Read article
AI & Agents
You Don't Have a Productivity Problem. You Have a Setup Problem.
Four hundred tasks across seven weeks. The sessions that produced client-ready work on the first pass all shared one trait. Not better prompts. Better setup.
Read article
AI & Agents
Three Files That Replace a Chief of Staff
Most people write their context files once and never open them again. The files work, but they deliver only 40% of what they could.
Read article
AI & Agents
The Prompt Is the Least Important Part
Everyone is optimizing the wrong variable. The people getting real results from AI wrote three plain text files and never looked back.
Read article
AI & Agents
Codex CLI Developer Guide
Safety-first Codex CLI setup for developers: install, secure profiles, automation tips, and configuration defaults for reliable day-to-day AI workflows.
Read article
AI & Agents
Enabling Autonomous AI Agents to Make Payments: Challenges and Solutions
Enable autonomous AI agents to pay safely—virtual cards, programmable wallets, crypto rails, and agent protocols, with real-world limits and safeguards.
Read article
AI & Agents
Mastering AI Model Fine-Tuning
Fine-tuning succeeds on data quality and evals, not hyperparameters. Shift from model tweaks to strategic AI consulting that builds lasting client value.
Read article
AI & Agents
Exploring Agentic Workflow Patterns
A visual guide to 11 agentic workflow patterns—network, hierarchical, sequential, parallel, router, and more—for building modular multi-agent systems.
Read article
AI & Agents
Mastering Cursor Rules: A Developer's Guide to Smart AI Integration
Learn Cursor Rules—global settings, .cursorrules, and .mdc patterns—that turn AI into a project-aware coding partner with agentic, contextual workflows.
Read article
Engineering
When Third-Party APIs Go Rogue, Strategies for Building Bulletproof Integrations
Harden integrations against flaky third-party APIs with timeouts, retries with backoff, rate limits, interface decoupling, and durable background jobs.
Read article
Engineering
Polymorphic Model Resource Finder in Ruby On Rails
Solve polymorphic association URL generation and parent lookups in Rails with a reusable resource finder pattern for cleaner, safer model code paths now.
Read article
Engineering
Publishable Enum in Ruby on Rails with PostgreSQL
Add publishable status to Rails models with PostgreSQL enums—migrations, scopes, and safe state transitions for reliable draft-to-published workflows.
Read article
Life & Work
9 Great Questions to Ask During the Hiring Interview
A guide to essential interview questions that help assess candidates effectively, covering motivation, cultural fit, self-awareness, and potential red flags
Read article
Life & Work
The Practical Guide to Gmail Productivity
Practical Gmail workflows, filters, labels, and UI tweaks to reach Inbox Zero faster and cut daily email time by ~70% without losing team responsiveness.
Read article
Engineering
The Ultimate Guide to Dokku and Ruby On Rails 5
Step-by-step guide to deploy Ruby on Rails 5 on Dokku, a self-hosted Heroku alternative—apps, databases, SSL, zero-downtime deploys, and production tips.
Read article
Life & Work
5 must have actions to make your emails more trusted
Five must-have steps before sending from a new domain or ESP: set SPF, DKIM, and DMARC, then verify deliverability so your email stays trusted in inboxes.
Read article
Life & Work
7 web tools for Perfect Naming for your Startup or Products
Seven web tools to find, check, and validate startup or product names—domain availability, trademarks, brand checks, plus creative naming inspiration.
Read articleKeep thinking with us
Practical AI ideas, delivered where you already are.
Get occasional field notes on choosing models, building useful AI workflows, and making better decisions with the tools.
Keep me posted by email
No fixed schedule. We will write when there is something worth sharing.
Confirm once by email · Unsubscribe anytime.