Signals

* Paul: The plumber who left school at fifteen runs his business on AI. The speaker after Paul at a Denver talk dropped out of an American high school, joined the military, learned plumbing as a trade, and stood up a business whose entire back end runs on AI: the right part to the right plumber at the right time, invoices, cash flow, the whole hairy mess Paul says he is glad is not his job. If a firm with five thousand plumbers had paid someone five million dollars to build that, it would be a nice story. This man built it himself. Sam’s point back: that firm would never have hired him. The bootstraps were built from the bottom up by someone motivated enough to make it happen and who did not know not to. If this lifts the bottom of the pyramid, the guy in Uganda spending weekends in Claude Code, then it is a technology that changes the human race, and that is the signal Paul says inspired him.

* Sam: Buyers are settling on good-enough models and going open. Paul’s hypothesis first: for any knowledge work in business the models have been more than good enough for a long time, gains are at the margin, what matters is the rest, the harness. His Hermes setup has a model switcher and will route work to Qwen or Kimi at a tenth or a twentieth of the price. So how does anyone make money out of Fable 5.1 when most people’s use cases could run on Kimi? There is no money in models. Sam’s signal is that the revelation is landing: in discussions he is in, organizations are saying we have (or are now standing up) the one that is good enough for us, and switching to an open-source model so they are not sharing all their information, because the data agreement on the frontier model is one many institutions cannot sign for legal reasons.

Subtractions

* Paul: Stop waving through work you did not read. Asked how a client would know AI adoption is going well, Paul says it is undiscovered country, then answers from his own practice: what are the behaviors, and how do you verify the work product AI generates? When he started, six pull requests waiting on GitHub got yeah, yeah, yeah, merge, merge, merge, and he ran without permissions because he has attention issues. Now he has to discipline himself to read code he does not fully understand, because he is using these tools so much he cannot give everything the pass. The tendency is human: that looks pretty good. A friend used to say “good enough for government work.” Think about it hard and it is kind of disgusting: it means this is not important enough for me to devote time to verifying it. That thinking is terrible for the age of AI. If he writes something himself he knows there will be no ridiculous mistakes in it. He does not know that about AI, and he spent the better part of a day rewriting what Claude wrote from his own markdown.

* Sam: The model is no longer the decision. Build the stack that is good enough for the work you are doing by how you divide up the work and, more importantly, the guardrails you place around it at the harness level. That is what constrains behavior and makes it trustworthy. It is not the model, it is what goes around the model. No different than with people, which is where last week’s episode ended: managing agents is much more like managing people than anybody is comfortable with.

Notes

The episode opens cold on Paul drowning in his own agents: they pump out so much that he is the one who has to verify it all, and a day of his week goes to validating what they did overnight. Then it backs up into the evidence problem he has been poking at since The Science of Organizational Change. The famous study proving change management makes you 60% more likely to succeed asks people how the change management was and how successful the change was, then correlates the two. Nobody has ever been in a project where the project succeeded and the change management sucked. Paul’s own best result, a KPMG culture change that took a group from the lowest revenue per partner to the highest, ran alongside a new CEO reorganizing, hiring and firing, and Paul cannot draw a causal line from what he did to the turnaround, though the CEO would say it was instrumental. No client has ever asked him whether he had evidence for what he was saying. Evidence-based medicine did not get its first paper until the 1990s and a surgeon told him in the 2010s that his department did not adopt it because it prevents innovation. People are perfectly capable of ignoring evidence when it is in their interest to do so.

That is the decision question in its old clothes. Sam’s question for AI adoption is the same one: organizations measure visible activity, policies, committees, training, tool availability, and the decision requires evidence of changed behavior or outcomes. Paul’s answer is that for most knowledge workers the agentic revolution started in February, nobody has written the book on running a multi-agent operation, and he is the glue: every finished task costs him half an hour to two hours to verify, approve, and revise, and with five or ten agents running he is drowning in open loops. The verification load is where the evidence lives, and it is also where “good enough for government work” does its damage.

Sam’s own harness answers Paul’s question about how to review without the reviewer producing the same bloated prose. One AI writes the spec and the test; a panel of AIs competes for the work by saying who can do it best and why; the winner runs it and cannot validate its own output, the spec writer does; a third model logs everything, and they are all from different model families so they cannot bluff their way out. That is Harnessie, Sam’s open-source project, named on tape in the course of the conversation; nothing was paid and it is not a sponsor segment. Paul’s response, having said the last thing he wants is another tool in his stack, is that this is frontier stuff.

The back half is management. Paul reaches for the One Minute Manager four-box: motivation on one axis, competence on the other, and with a motivated, competent person you say go do this and tell me when you are done. Ten years of shared context lets you say make me the deck for Friday; open a fresh Claude Cowork and say the same thing and you get the average of all decks. Sam’s version is that the expression of intent should be intentful, and that what he is building toward is aggregated intelligence: not people first versus AI first, but the maximum intelligence you can bring to a problem, with intent honed enough that people and machines pull the same way and you can tell when they are not. When Sam says the thing that keeps people in a predictable set of behaviors is the environment around them, not the request or the instructions, Paul names it: that is nudge theory, context and environment predict behavior better than motivation, purpose, or skills, and it was the whole contribution of that school of change management. The good news is that if you have ever managed anybody, you have a leg up on managing agents. Paul’s confession closes it: he is a good leader and a bad manager, his son came to work for him for two weeks and needed hourly check-ins Paul never gave, and that failure mode exists for agents too. Send one into virgin territory with come back at six o’clock and you burn tokens or introduce risk.

Paul would not put anyone on a change team who is not really good with AI, and he is about to sell some big change management work. He knows maybe a dozen people who can talk credibly about org change and leadership and know what they are talking about with AI, and thinks that spot in the middle ought to be the most valuable coin of the realm, because the difficult stuff is the workforce stuff and everyone is spending the money on technology.

About

Every Wednesday, we livestream. Every Friday, an episode like this one. Stay tuned for issue 67 of the newsletter on Sunday. Check out https://sigsub.show/ for more.

Signals & Subtractions is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Signals

* Lee: Build the test before you trust the analysis. When something doesn’t work as expected, build a test around how the user will actually use the solution, whether that is a trench or a learning management system. The most powerful tool he has is real data, not assumptions, not ideas, not how it should work. His move this week: more informational interviews, working the question at the level of why. Not how, not what, not when, not how fast. Why is this what I’ve decided to do?

* Sam: The agents found the product before the humans did. Months of long sales cycles pitching everyailaw.com to law offices that are wary of AI, then a spike of inhuman activity in the MCP analytics: agents had discovered a product that breaks human laws down into machine-readable text, exactly the thing they have almost nothing like. Agents have had their own wallets since April, there are more bots than people on the internet, and the surface they found does not even have a payment gateway yet. The signal is the rethink: redesign the business for an agentic market that was not a market six months ago.

Subtractions

* Lee: Remove words the problem statement to find the solution. From the genius bar: 95% of the time, the answer was in something the person said when they first sat down, the detail you glossed over because you had already decided it was a hard drive problem (”Tommy put new RAM in my computer a while ago”). Break the problem to its simplest state, dig a hole at each end of the pipe, and ask whether you need what you are doing at all, before you argue specs on a trench you don’t need.

* Sam: Subtract the word “agent” you don’t need it. Drop the label and get close to the work: here is the input, here is the output, what makes it flow smoother? Then upgrade the questions the label was hiding. “How do I stop AI from hallucinating” has no answer, because on the AI side there is no mechanistic difference between hallucinating and answering; “how do I manage the context it has” does. “Why isn’t the agent doing what I told it” becomes “how do I scaffold the guardrails around the agent rather than inside it.” Bounds you need kept are architecture, not instructions.

Notes

The episode runs on one move made three times. Lee digs a one-foot hole at each end of his water line and cancels a $4,000 trench: input and output validated, nothing in the middle needs fixing. Claude asks him why he is using an LMS at all when what he described was a knowledge base, and weeks of forcing the wrong tool evaporate. And Sam finds inhuman traffic in his analytics and realizes the right response is not to fix the funnel but to requestion the business.

The single loop learning fixes the mistake. The double loop learning fixes the question.

The connective tissue is Lee’s split half troubleshooting from his Apple genius years, written originally by Steve Wozniak: simplify and reduce the variables before attempting to solve, test first, verify the physical layer. Sam ties it back to Douglas Hubbard’s uncertainty-reduction method from Ep 7, the same Bayesian instinct wearing different clothes. The back half turns to agents: the Fortnite Darth Vader that two 16-year-olds prompted past its guardrails, and the cold-open rule that if you give an agent information and tell it not to share it, it will share it. Agents are not secured in and of themselves; security is a layer you put on top. If you want strong bounds, build them as architecture, with instructions as commentary.

Lee references Sam’s blog post “The Question Is the Answer” twice. The episode is effectively that argument field-tested against plumbing, WordPress, and an agentic sales funnel nobody planned for. Along the way testing stays human work throughout: don’t take AI’s word for what AI did. Take the time to check.

Signals & Subtractions is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Everyone has ideas now. AI made sure of that, and it quietly repriced the thing that used to separate people at work: not having the insight, but getting the right person to pay attention, reconsider, and decide.

Cohost JD Dillon takes the chair for the first time, and guest Josh Felix brings the story that anchors the hour: a three-and-a-half-minute layoff, 64 recorded conversations run through an AI to find the trend, and five job offers built on a website instead of a resume. Three signals worth watching and three subtractions worth making: delete the resume, stop chasing scale, and stop shopping for the best model. JD lands it on Trader Joe’s, a store that refuses the self-checkout and owns the most profitable square footage in the grocery business, because it is ruthlessly clear on what it is.

JD is the author of The Frontline Enablement Playbook; his book gets a plug in the show, unpaid and disclosed. Josh is senior director of professional services at Oxford Global Resources.

Full transcript and links: https://sigsub.show/episodes/ep-008/

Watch the video version on YouTube:

Show notes:

* Josh Felix on LinkedIn: https://linkedin.com/in/joshfelix/

* TheFutureWithJosh.com, the site Josh built as his resume

* Resume.Sam-Rogers.com, another similar example

* JD Dillon on LinkedIn: https://linkedin.com/in/jddillon/

* The Frontline Enablement Playbook:

https://frontlineplaybook.com

* The Modern Learning Ecosystem:

https://jdwroteabook.com

* Harnessie, the free & opensource multi-agentic harness mentioned on air

* Episode page and transcript: https://sigsub.show/episodes/ep-008/

Signals & Subtractions is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Douglas Hubbard invented Applied Information Economics and wrote How to Measure Anything. In 2016 I asked him how you would measure something everyone agrees is unmeasurable. He asked me how I would measure collaboration, and I gave him the answer everybody gives: count the messages.

Ten years later that is still the answer, it is still wrong, and now it is on a dashboard with the word AI at the top of it.

This episode is an interview from my archive, recut with a new introduction and close. It predates ChatGPT, copilots, and any AI budget anyone had to defend to a board, which is the reason to play it now rather than a caveat about it. Nothing in it needed updating.

In this episode:

* Why there is no such thing as a statistically significant sample size

* Why you have more data than you think and need less than you think

* How to define what you are measuring by the decision it changes, not the thing you can count

* Why refusing to price a human life just means pricing it badly and in secret

* Why you are worse at confidence than you think, and how half a day fixes it

Douglas Hubbard: howtomeasureanything.com

Audio only this week. Full transcript and every surface: https://sigsub.show/episodes/ep-007/

Signals & Subtractions is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Guest Michael Simon on what is actually in force: most of the attention went to the timelines that moved, while a set of chatbot obligations came into force on schedule across US states and beyond. Federal and nine states in detail, the EU AI Act and California around them, and what an operator does about it on Monday.

You can find Michael Simon at https://linkedin.com/in/michael-simon-9400014 or via his law firm https://www.lawplusdata.com/

Also mentioned in this episode:

https://everyailaw.com/

https://obligationfirst.org/

https://aiincidentlaw.org/

For the full notes, transcript, and additional resources, visit: https://sigsub.show/episodes/ep-006/

Today marks the launch of the Signals and Subtractions podcast proper. We wanted to get a few episodes under our belt first, and this audio trailer makes it official.

So follow where you please, and if we’re not there yet just ask and we’ll add your favorite podcatcher too.

Apple Podcasts | Spotify. | Pocket Casts | Substack Podcasts

Nothing Was In The Water — Limited Edition Jonathan's piece, the premise of the episode

Against Claudefishing — Chris Best's case for the feature

Writing Quality After AI — Sam's article on origin, process, and quality as separate properties

Limited Edition Jonathan on Substack

Canonical episode: https://sigsub.show/episodes/ep-004

Two voices this week: host Sam Rogers and guest Sabino Marquez. Sabino is twenty years a CISO, adviser to CISOs, and for two years the Canadian ambassador to the Global Council for Responsible AI. Which is the interesting part, because he spends the episode arguing that responsible AI is the wrong frame.

The arc runs from a conversion story to a rebuilt first question. Sabino came up in the nineties when protecting your technology was protecting your advantage, watched Sarbanes-Oxley put IT under the CFO after the turn-of-the-century scandals, and found that the more compliance he did the less safety he could prove. His diagnosis is that the business never saw a defender; it saw a security-shaped answer to a security-shaped problem standing between it and the dollar it wanted. That is how a CISO ends up accountable for every trust-creating practice in the company without being empowered to run any of it, and how the trusted-advisor frame becomes a trap: you are important, you are allowed in the room, and you do not have a number to deliver, so you are advice, and advice gets overruled.

The evidence-theatre block is the one to watch. Selling to the Fortune 50 means diligence teams that count every armpit hair one by one, so Sabino booby-trapped his own PDFs with a tracking pixel to find out whether anyone actually opened the documentation his company spent six figures a year producing. Mostly they did not, and his conclusion is not that buyers are lazy: handing someone forty documents is asking them to assemble a puzzle without the picture on the box. Data makes people feel something. The IT reviewer, the lawyer, the insurer, the regulator and the procurement lead all need a different story told from the same evidence, which is what it means to run the company as a factory that produces documented proof that value is safe in your hands.

The constructive turn is that this pays. Sabino puts it at roughly half a billion dollars more across ten years and multiple companies: higher exits, bigger rounds, higher ACVs, faster sales at lower cost. His closing claim is that safety becomes the number one buying criterion, ahead of price and quality, because we have entered an era of unrestrained velocity where the benefit of speed is outrunning the explosions you set behind you.

The value inventory is the part most companies fail cold. Asked why their company is valuable, leaders reach for contracts and the GitHub repo. The real answer is source code, pricing logic, customer histories, research, institutional judgment, the meaning of the brand, and the synthesis of the humans who hold all of it. Those are not raw material to be transformed into value. They are the value. You cannot prove an AI is safe for value you never identified.

This week’s links:

* OpenAI says AI models escaped control and hacked Hugging Face (Fortune): the day-of story behind Sam’s signal

* The Global Council on Responsible AI

* Trustable.tv & TrustClub.tv: Sabino’s consulting practice and his ongoing blog, where most of the system is published as essays and instruments, aimed at people who do trust work for a living

* Sponsored by Trustable, which builds instruments for measuring current trust value and designs operating conditions that compound it.

This week host Sam Rogers streams from Cancun with cohost Lee Rodrigues and guest Christine Rodrigues, partners in life and in business, to work through what AI did to the job hunt from every side of the table: the seeker, the hiring manager, and the recruiter. Canonical episode: https://sigsub.show/episodes/ep-003

The throughline: AI raised the quality floor, which also raised the noise floor. When everyone can generate a beautiful, perfectly-matched resume, the beautiful resume stops meaning anything, and the human parts of the process become the only real signal left. Christine calls it thought sovereignty, the discipline of keeping the original thought, the voice, the taste, and the judgment even while using the tool. Lee’s version is alignment across your surfaces; Sam’s is aiming for the tails of the distribution instead of the average.

The current-events beat lands on data hygiene: Lee’s example of the man whose AI chat logs ended up in a murder case makes the point that there is no attorney-client privilege with the tool on your phone, which is the same reason you never feed a pile of applicant resumes to a public chatbot. Jobs are human things. The move is not bigger numbers, it is the right numbers: a phone call, a warm relationship, a referral outside the herd.

The signals

* Christine: The resume stopped being a signal. A good resume used to signal a good candidate, the same way good writing signaled a good writer. Now anyone can upload a job description and generate a polished resume that matches it exactly, and everyone using the same tool on the same posting converges on the same words. The polish is an anesthetic: it looks great, which numbs you to whether the substance is there. What used to be signal is now noise.

* Lee: Are your resume, LinkedIn, and portfolio the same person? Lee ran black-hat thinking across his own three assets and Claude asked, are you sure this is you. They had drifted apart since 2015. Multiple signals that do not line up read as noise. Pick one focus and align every surface to it instead of wiggling all of them at once.

* Sam: AI writes you toward the average. AI is strong in the middle of the distribution and weak at the tails; the [Mount Sinai study](https://nature.com/articles/s41591-026-04297-7) found it good on common medical questions and worse than humans at the edges. An AI-drafted resume makes you look like the average of every resume out there, which is perfect if you want the middle of the pile and fatal if you want the top five percent. The memorable candidates are corner cases: the small-town newspaper editor who could actually write, the cruise-ship costume designer who could improvise a show in thirty minutes.

The subtractions

* Christine: Stop delegating the whole document. Do not push a button, generate a resume, and send it. Chunk the work, keep a master document you control, and transcribe the AI output by hand so your voice survives; her fingers will not let her type something she would not say. Every AI-written bullet is a story you will have to defend in an interview you never rehearsed.

* Lee: Unpublish the portfolio you will not maintain. If you are not going to keep it current, delete it, or rebuild it as one clean, simple, narrative page and cut everything unnecessary. It feels like spring cleaning, and a stale portfolio that contradicts your resume costs you more than no portfolio at all.

* Sam: Do not dump the resume pile into a chatbot. Handing a stack of a thousand resumes to ChatGPT to pick your five is illegal in most places, and it leaks data that applicants entrusted to you. Match the conversation to the channel, secure the traffic in between and not just the endpoint, and remember that local models keep the sovereignty in your hands.

This Week’s Links

* The Mount Sinai study on ChatGPT Health (Nature Medicine) — strong on the common questions, worse than humans at the edges

* Former NFL player consulted AI chatbot after murdering his girlfriend — Lee’s no-privilege example from the show

* Easy, Awry — Sam’s comedic essay on one-click job applications, receipts from nearly a year ago

* System2Focus — Lee and Christine’s workflow-based approach to the job search

* R&P Associates — strategic planning and marketing for nonprofits

Sponsored by System2Focus, Lee and Christine’s workflow-based approach to the job search. Next week: Sabino Marquez on the CISO perspective in the age of AI. Cohosts JD Dillon and Markus Bernhardt join the rotation next month.

Want to bring your own signal and subtraction? Find yours. Sponsored by System2Focus.

Canonical episode: https://sigsub.show/episodes/ep-002

This episode sets the ground rules: what this show is, what it isn’t, and what we actually mean when we say signal and subtraction around here. By the end, you’ll be able to find your own signal and subtraction, with or without AI.

First, the news:

* Anthropic’s Jacobian Lens (JSpace): a new paper and open tool that lets you peer into a language model’s “subconscious,” the precursors that come before the thinking and the output tokens. Free to play with on Qwen and Gemma, with a GitHub repo. Sam’s take on why it’s news: not that AI got smarter, but that the black box just went a little transparent. Read the post from Anthropic or watch their video here.

* Fablemaxing before the window closes: Fable 5 is currently the most powerful commercially available model, and Anthropic’s billing change now lands Sunday, July 12. If you haven’t hit your token limit yet, there’s a few more days.

* Microsoft’s Frontier Company: a $2.5B, 6,000-person AI deployment unit embedded with enterprise customers. More evidence that we’ve moved from 2025 pilots to 2026 production. It’s infrastructure time.

Then, the method. Sam works through a real example live, by hand:

* What makes a signal: recent, specific, first person. If a stranger could have posted it on LinkedIn, it isn’t yours yet.

* What makes a subtraction: something you actually stopped, killed, or refused. Past tense. Made, not planned. Not advice for other people.

* The analogy layer: recurring meetings as for loops that just count, versus do-while loops that know when they’re done and hold their own exit condition. Convert the for loops into do loops.

* Make it testable: open your calendar, look at each recurring event, and ask two questions. What’s its exit condition? Who checks it? No exit condition and no checker means you’re running a for loop for somebody else. Give it a return statement this week.

Want the home game? Grab the AI prompts at SigSub.show/find-yours and play with your bot friend anytime.

Livestreaming Wednesdays. Polished podcast/video Fridays. Newsletter recap Sunday.SigSub.show

Signals & Subtractions is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Thanks to the sponsor of this weeks’ show: Harnessie.com! Put AI to work without giving up control with this free and open-source, multi-agentic harness based in agentic collaboration to enable human ownership.