This website uses cookies

Read our Privacy policy and Terms of use for more information.

Issue 11 · The Slot

Hi friend,

Two changes before anything else. This letter is weekly now, every Monday, instead of turning up twice a month. And it looks different because it is different. The sections are windows, the roundup of new product launches is gone, and what replaced it is Last Tool Standing, a public directory of what holds a slot on my desk and what got cut. The news stayed. The things you can actually apply to stayed, and so did the events worth the drive.

Last issue I asked what your tools are never allowed to touch. That question is still open, and I would still like your answer.

The rest of this issue is built around one rule I have never written down for you, because I spent last week somewhere that convinced me it was time.

🔨 Built Wrong

You are collecting tools when you should be building skills.

I taught in San Francisco last week, and the class went well, but the thing I brought home happened on the street. Where I live you can go a month without seeing a single advertisement for AI. In San Francisco I could not go a block. By the second day I stopped being impressed and started noticing what it would feel like to live under it, because if AI has been good to you those ads read like progress, and if AI is the reason your team got smaller last year, every billboard is a reminder you cannot opt out of on your commute.

That is the soil your tool decisions grow in, and it explains something I see in almost everyone I teach. When people feel behind on AI, they add tools, because adding feels like progress and costs one signup. But a tool without the skill underneath it just moves the anxiety somewhere new. Now there are four tabs doing one job, nothing was decided, and the feeling that you are behind has four subscriptions attached to it.

The pattern that holds is the reverse. Skills first, then one tool per job, chosen on purpose. Skills are the only part of your stack that survives a layoff, a price hike, or a product shutting down, and they are also the only part that transfers when you build something of your own.

I get to say this because I did the wrong version myself. I teach this for a living, I think AI is genuinely great, and I still spent July with three tools that all write files, telling myself the overlap was flexibility. It was not flexibility. It was a decision I was avoiding, and the next section is the rule I used to finally make it.

⚡ One Slot

The One Slot Rule

Every job on your desk already has something doing it, and a new tool does not get to move in alongside. It has to take the slot. That is the entire rule, and it sounds harsher than it is, because the moment you say it out loud, the tool argument you have been having for three weeks tends to resolve in about a minute.

It works because it changes the question. You stop asking which tool is better, which nobody can answer, and start asking what the job is and who currently holds it, which you can answer immediately. So you go in this order. Name the job in the plainest words you have. Ask whether the new thing can do that job, not whether it can do more. Ask what it touches that would be hard to undo. Then ask what would have to be true before you would let it run without you watching. If the new tool cannot take the slot cleanly, it does not get a slot, and you have saved yourself the afternoon.

One placement note so nobody mistakes this for a philosophy. The Pyramid is how you hold the work, ESP is how you do the work, and One Slot is only how you choose what you do it with. It sits under ESP, not beside it.

Starting this month the rule also has a public home. Last Tool Standing (https://charmthirteen.com/last-tool-standing) is a directory of the tools holding slots on my own desk, one row per job, and the verdicts are Still Standing or Cut. The Cut entries publish too, because a directory that only says nice things is an ad.

🌎 In the News

Your view count is about to go up without you doing anything better.

From today YouTube counts a view the moment the first frame plays, across long form, Shorts and live. The stricter measure does not vanish, it moves into Analytics as Engaged views, and your earnings and your Partner Program eligibility keep running on that engaged number rather than the public one.

So August stops being comparable to July on every channel at once. If you track this anywhere, that column needs a date and a label rather than a silent jump.

One correction while everyone repeats it wrong. The thirty seconds people keep citing was never published as policy, so it is folklore YouTube declined to carry forward, not a rule it retired.

✅ Before You Ship

Find the error before you use the output.

Take the last thing AI produced for you, an email, a summary, a plan, and find one error in it before you act on it. Not read it over, find one. A wrong number, an invented name, a date that slipped, a confident sentence with nothing under it. If you cannot find one, pick the single most consequential claim and verify it against a source.

This is the skill that separates people who use AI well from people who forward its mistakes, and it takes about ninety seconds once it is a habit. It is also the exact skill we drill live in the session below, on your own work instead of my examples.

💻 Prompt for Productivity

Run the One Slot Rule without letting yourself skip to the verdict.

❝

You are helping me decide whether a tool earns a slot. The job is [the job, in the plainest words you have]. The tool doing it now is [tool]. The tool I am considering is [tool]. Ask me these four questions one at a time, and do not move on until I have answered. What is the job. Can the new tool do that job, rather than more. What does it touch that would be hard to undo. What would have to be true before I let it run without watching. Then tell me whether the new tool takes the slot, and name the one thing that would change your answer.

WHEN TO USE IT: the moment you notice two tabs doing the same job and a third one bookmarked.

WHAT IT DOES: it holds the questions in order and refuses to jump to the answer, which is where tool decisions usually go wrong. One at a time matters, because four questions asked together get answered as one paragraph, and a paragraph is easy to hide in.

TIP: answer the hard to undo question out loud. If you cannot say what the tool touches, you are not ready to hand it the slot, and that is your answer.

— Lex

🗓 Open Calls

September 15. Plug In Ventures is running a free pre-seed accelerator and pushing it at Kansas City founders building AI-enabled products. It is self-paced and virtual over about two months, nine learning modules and four moderated sessions, with mentors and Entrepreneurs-in-Residence, and it is not cohort-based so you start when you start. Onboarding begins around November 15. Be clear-eyed about what it is, because their own FAQ says there is no direct fundraising support, so this is curriculum, people, and a room, not a cheque. They state a focus on underrepresented founders. Apply here (https://pluginventures.co/uncategorized/faq-for-ai-apps-pre-seed-accelerator/).

September 14, 5pm Pacific. The Agents for Humans hackathon, run by AWS on Devpost, wants an agent built with the Strands Agents SDK that takes something people actually deal with and handles it end to end instead of chatting about it. Forty thousand dollars across four prizes, ten of it the grand prize. The part that matters for this list is the eligibility, because companies and teams can enter and not just individuals, which is rarer than it should be. New projects only, built inside the window, with a public repo and a video of five minutes or less. There are fifty dollars of AWS credits if you ask by September 11. Rules and entry (https://agentsforhumans.devpost.com/).

September 17, 11am Central. Clock In to AI, my one hour live working session, runs its first open seat. You bring one real task from your week, we finish it with AI together, you learn to catch the places AI gets it wrong, and you leave with the templates and guides. A seat is $95, registration is on Luma and subject to my approval, and the seat credits toward the full Working 9 to AI series if you keep going. Register here (https://luma.com/u4nc3gjy). It repeats the third Thursday of every month, so October 15 and November 19 are already set. The longer story of why I built it is on the blog (https://charmthirteen.com/blog/why-i-built-clock-in-to-ai).

After September 11. Registration opens for Claude Loves Lovable, the five session vibe coding series. Reply if you want to hear first.

🔭 Next From Me

Systems in Practice, the quarterly meetup, becomes a podcast, and the first episode is close.

💌 One Ask

What job on your desk has two tools holding it right now? Reply with the job, not the tools. I am collecting these for Last Tool Standing, and yours might become a row.

Reply

Avatar

or to participate