Notes on working with AI
Field notes from teaching Claude to real teams. What works, what breaks, and what we keep relearning.
Featured · 8 min read
No independent test shows Superpowers, ECC, Spec Kit or GSD beating plain Claude Code. What the evidence says, and which one to try for which problem.
3 min read
The slow part of building with Claude is no longer producing the work. It's reviewing the plan. A plan shaped as decisions you answer gets reviewed; a long text file gets skimmed.
3 min read
When Claude builds a video in code, every frame is a function of time. That is why each frame can be checked, and why a storyboard before rendering saves hours.
3 min read
An agent you can't see is one you can't trust, and one you can't grade can't improve. A trace and a fixed test set close most of that gap.
3 min read
Pre-trained forecasting models can predict your data without training on it. The value is in the error band and the backtest, not in the line.
3 min read
Many AI projects are classic prediction problems with a table of past data. A language model is the wrong tool for them, and Claude is most useful as the analyst who builds the right one.
3 min read
Most work in an AI workflow is small, routine calls. Paying top-model prices for them is the main waste. Use a big model to plan and small models to do the volume.
3 min read
A language model does not remember your conversation. It rereads it. That one fact explains why a three-hour chat is slower, uses more of your limit, and loses early details.
2 min read
Everything always loaded in Claude Code is re-read every turn. Audit your instructions, skills and connectors like a subscription bill.
4 min read
A company AI that answers every question has a design defect. The fix is a refusal threshold, cited sentences, and a test set that includes questions it must decline.
4 min read
An agent is data access, written rules and a scheduler. The design choice that matters most is keeping read permissions apart from action permissions, then proving the rules in a dry run.
3 min read
Calling a language model autocomplete is accurate about its objective and wrong about its capability. The pipeline explains why, and what it means for how you use it.
3 min read
A model that returns a probability over fixed options routes work better than one that writes a persuasive paragraph. Calibration, latency and cost per decision decide which model you need.
3 min read
A language model knows a product as an average of what the web says about it. Skills let the vendor, or you, hand it the actual procedure instead.
4 min read
Automated sorting stalls when the rules are written once. A cascade with a confidence bar turns every escalated case into an edit to a written rubric, which keeps the improvement auditable.
2 min read
Handing Claude a whole job is the part of the demo that gets a reaction. Turning that job into a scheduled task is what gives you the time back every week.
3 min read
The Claude chat can take on a whole job. It works through the steps, picks the tools and hands back finished files. Here's how to brief it and stay in control.
3 min read
A phone propped on a coffee mug, three folders handed to Claude Code, and a full ad edit came out the other side. The setup is the part worth studying, not the polish.
3 min read
Bring real job descriptions, a policy document and an actual hiring spreadsheet into one Claude workshop, and here's what comes out: a job post in minutes, an onboarding kit ready for day one, and an instant answer to every policy question.
3 min read
A stuck PRD, a pile of unsummarized user interviews, a metrics dashboard nobody checks. Here's where a Claude workshop for product teams actually saves hours, not another slide about AI.
3 min read
Not another chatbot that reads a pitch deck faster. The three skills a VC fund builds first with Claude, and why the order you build them in matters.
3 min read
Ask Claude to check today's price and it might tell you it can't, its knowledge stops at training. A connector fixes that, but only if you pick the one built for the task you actually have, not all five at once.
3 min read
The cost of having Claude review a video doesn't track its length. It tracks how often the picture changes, and that flips which recordings are actually worth watching in full.
4 min read
Every MCP connector used to work like a call center that forgets you the second you're transferred. That's changing, and it means more of the software you already use is about to plug into Claude.
4 min read
A list of GitHub repositories promised to make money on their own. A closer look at what two of the five can actually do for a small business, and why the other three need an engineer first.
4 min read
An instruction file that grows one frustrated rule at a time eventually fights itself. The newer models do better with a short one that trusts judgment instead.
4 min read
A calendar slot is a fact you can check. A customer's judgment call isn't. The gap between them is the whole question worth asking before you automate anything in your business.
3 min read
Split work across specialist agents and each one gets sharper at its slice. Add enough of them without enough distinct work to justify it, and you've built a meeting, not a team. The rule is the same one that decides when an event planner needs a lighting specialist.
7 min read
Every enterprise deal hits the same moment, when the CFO or the general counsel asks whether it's actually safe or legal to put company documents into an AI tool. The straight answer holds up better than the reassuring one on the slide.
6 min read
Every team we train eventually asks which tool to standardize on, Claude or ChatGPT. There are real answers on long documents, working inside Excel, and privacy, but the difference that actually decides outcomes isn't the tool, it's how the task gets built around it.
3 min read
Three saves landed in the same folder this week, and none of them taught a technique. Together they say something about what capability looks like now. It arrives in bulk, and installing it is not the same as knowing which three pieces you will actually use.
5 min read
You searched for a price and found a form instead. We won't publish a number either, and here's why, plus what actually drives the cost of a corporate AI workshop: group size, format, how much it's built around your own documents, and what happens after.
3 min read
Most teams that get an agent working never get past the demo, and the reason is almost never the model. The real question is whether you can ask it, each evening, what it did, and get an answer.
4 min read
Everyone is posting diagrams about loop engineering like it's a new discipline. It isn't. Here's the actual anatomy, the one idea in it worth taking seriously, and the honest test for whether you should build one at all.
3 min read
We run enough scheduled Claude Code sessions now to actually watch the token bill, and the biggest driver of it has nothing to do with how much work got done. It's a caching mechanic almost nobody explains, and it can turn one word into a real charge.
3 min read
After enough workshops, the same three habits show up in almost every new group. None of them are about technical skill. They come from a mental model built by years of using search engines and email.
1 min read
MCP is the protocol that lets Claude connect safely, read-only, to a business's real systems, turning a single chat into a working team of agents.
Claude training for teams
This is the material we teach in workshops, at your office or online. Tell us what you're trying to solve, and we'll come back with a plan.