Before a workshop or post: scan the topic section. Prefer entries with real "Landed?" evidence. Read the weak spot before you rely on an analogy, because that is where a sharp audience member will push.
After you use one: update Tried where and Landed? with what actually happened. Useful evidence includes a laugh, a question that shows people got it, an attendee repeating it back, or comments and DMs on a post. "Unknown" is a fine answer. A guess is not.
When you hear a new one: text it to Ray. It goes in with its source and first-seen date.
Sources: iMessage dates refer to the Ben and Ray thread. Vault paths are relative to ~/rdco-vault/. "Ray" means Ben's AI assistant proposed it.
Landed? discipline: an opinion from Ray is not evidence that an analogy landed. Only Ben's reaction, an audience reaction, or post engagement counts.
Status as of 2026-10-15: no LinkedIn draft in 01-projects/personal-brand/ has been published yet. Every "LinkedIn post" entry below is a draft placement, not a test.
Teaching
Cake and icing
Analogy: Every facilitator gets the same cake (the lesson plan). Each one adds their own icing: their personality, and how they relate the material to the client.
Explains: Standardized curriculum with room for each facilitator's style.
Source: Colleague (Andrew)
First seen: 2026-10-15 (iMessage)
Tried where: Workshop (used inside the facilitator team)
Landed?: Unknown. Ben asked whether it is any good (2026-10-15). There is no audience test yet.
Weak spot: Icing sounds decorative. Relating the material to the client is how the lesson lands, not a garnish.
Topic: teaching
Same recipe, local ingredients
Analogy: "Same recipe, local ingredients."
Explains: The same alternative to cake and icing. The method is fixed, and the client material is what makes it work.
Source: Ray
First seen: 2026-10-15 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: It drops the facilitator's personality, which is half of what Andrew's version covers.
Topic: teaching
Recipe ingredients at the top of a tutorial
Analogy: List the concepts a tutorial will apply at the top, like the ingredients on a recipe. The framework stays the same, the data for your situation differs, and you reach decision Y or Z.
Explains: How to structure a teaching unit so learners know the toolkit before they use it.
Source: Ben
First seen: 2026-04-05 (iMessage)
Tried where: iMessage (product idea for Data Dots tutorials)
Landed?: Unknown
Weak spot: Recipes promise the same dish every time, but analysis tutorials are supposed to lead to different decisions.
Topic: teaching
Watch one, do one, teach one
Analogy: The medical-residency progression: watch one, do one, teach one.
Explains: How a skill moves from observing, to practicing, to teaching. This is a candidate core belief for his personal brand.
Source: Ben (a medical-training saying). Diamandis's law "do it by the book, but be the author" is mapped to it in 01-projects/personal-brand/2026-10-04-diamandis-laws-and-data-for-good.md:13.
First seen: 2026-10-04 (iMessage)
Tried where: Not yet (brand core-belief draft)
Landed?: Unknown
Weak spot: It implies one repetition is enough, but most agent skills take many reps before you can teach them.
Topic: teaching
A vocabulary for taste (art history / "Data Dots for design")
Analogy: "I like different styles of art but lack the vocabulary." Collect a mood board and tag the design concepts, like Data Dots for design.
Explains: You can't direct or critique work (or an agent) until you have names for what you like.
Source: Ben
First seen: 2026-04-23 (iMessage)
Tried where: iMessage
Landed?: Ray agreed. It turned into the reference-sheet design workflow. There is no audience test.
Weak spot: Naming a style is not the same as having taste. A vocabulary can make you sound sure of a choice you haven't earned.
Topic: teaching
Grandma's cookbook (data intuition)
Analogy: Analytics know-how is passed down like grandma's cookbook, as hand-scribbled recipes inside each company. There are no public "recipes" the way cooking has YouTube.
Explains: Why data intuition is hard to learn outside an apprenticeship.
Weak spot: A captured idea can still go stale, so writing it down is not the same as keeping it alive.
Topic: teaching
Four kinds of ambition: speed, altitude, depth, range
Analogy: "I think about ambition in these four ways: speed, altitude, depth, range." Altitude means asking for the goal behind the task. Depth means using an expert's real tools. Range means assembling a crew of roles.
Explains: Ways to ask more of a model than "do this task faster."
Landed?: It landed with Ben as a literary device, because "people can visualize each of those aspects" (2026-10-15).
Weak spot (Ben's critique): It's hard to go high altitude and deep on something at the same time. The two dimensions pull against each other, which could be a post on its own.
Topic: teaching
Skills
Skills as text expanders
Analogy: If you reuse a long prompt, put it behind a skill. The slash command then works like a traditional text expander.
Explains: What a Claude skill is, for people new to skills.
Source: Ben. It first shows up in his workshop debrief notes, and it is unclear whether he or an attendee coined it.
First seen: 2026-10-09 (iMessage)
Tried where: Workshop
Landed?: Unknown. It probably lands best with people who already use a text expander.
Weak spot: A skill does more than expand text. It loads only when relevant, carries judgment, and can bundle scripts and files.
Topic: skills
A recipe card the agent pulls off the shelf
Analogy: "A recipe card the agent pulls off the shelf when the dish comes up."
Explains: Skills as on-demand, situation-triggered instructions. This is the alternative to text expanders.
Source: Ray
First seen: 2026-10-15 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: It drops the "you trigger it yourself" mental model that makes text expanders easy to grasp, and it may collide with the Agent Brigade kitchen vocabulary.
Topic: skills
MCP is the kitchen, skills are the recipes
Analogy: "MCP is the professional kitchen, skills are the recipes."
Explains: The runtime split between tools/connectors and know-how.
Source: Article (Anthropic skill guide), compared in 08-tooling/2026-06-30-anthropic-skill-guide-vs-brigade-comparison.md:19
First seen: 2026-06-30
Tried where: Not yet
Landed?: Ben and Ray dropped the comparison as noise. "Kitchen" collides with the brigade metaphor (2026-06-30).
Weak spot: Two kitchen metaphors in one talk will blur together.
Topic: skills
Progressive disclosure: a pointer file with a shelf behind it
Analogy: Structure a context folder "like a skill with progressive disclosure": a main pointer file says when to read each deeper file, and those files load only when needed.
Explains: How to organize a large context corpus so an agent can navigate it.
Source: Ben
First seen: 2026-06-28 (iMessage). He reused it on 2026-08-16.
Tried where: iMessage
Landed?: Ben reached for it twice, so it is a stable internal analogy. There is no audience test.
Weak spot: It only works on people who already know how skills load, so it is circular for beginners.
Topic: skills
Soul file as a creative brief, not a call-center script
Analogy: "Give it a voice, not a script... Think of it like a creative brief, not a call center script."
Explains: Writing an agent's personality file as principles to improvise from, not canned responses.
Source: Article (unattributed AI-generated text that Ben pasted)
First seen: 2026-03-28 (iMessage)
Tried where: Not yet. It shaped Ray's own SOUL.md.
Landed?: Ben acted on it: Ray's personality file was built that way.
Weak spot: Creative briefs assume a skilled human fills the gaps. A model fills them with defaults.
Topic: skills
Agents
Agent Brigade (brigade de cuisine)
Analogy: Run multi-agent work like Escoffier's kitchen brigade. Tickets go on a rail, stations do narrow jobs, and an expo at the pass decides what goes out. Context bundles are "the tickets that go onto our rail" (2026-06-28).
Explains: Reliable multi-agent orchestration, positioned against agent swarms. It covers role separation, one decision point, and an audit trail.
Source: Ben. He developed it with another AI chat (2026-06-28) and refined it with Ray.
First seen: 2026-06-28 (iMessage)
Tried where: Graphics to communicate it internally to his org (2026-07-02), plus the RDCO plugin design
Landed?: Ben: "I'm pleased with how well this Brigade de Cuisine fits... Many people have worked in a restaurant... or have seen The Bear or Ratatouille" (2026-07-02). A reviewer critique he pasted called it "charming but heavy" (2026-07-05).
Weak spot: It comes with a vocabulary tax. People must learn Expo, Rail, Cellar, and mise before they can ask basic questions.
Topic: agents
Front of house: waiter, sommelier, set menu
Analogy: Gathering context is the waiter taking the order before hanging the ticket on the rail. Or it's a sommelier: who defines the menu?
Explains: The intake step that comes before agent work starts.
Source: Ben
First seen: 2026-06-29 (iMessage)
Tried where: iMessage
Landed?: Partly. It produced the "steward" and "menu" concepts. Ben later warned: "I don't want to build it out just to extend our analogy" (2026-07-03).
Weak spot: It overextends the metaphor. Each new kitchen role adds vocabulary without adding mechanism.
Topic: agents
Soft opening / ignition sequence
Analogy: Ben: a "soft opening" mode. The mise says whether the brigade is ready to work in your environment, the menu says what it can prepare, and service starts pulling tickets. Ray's variant: menu = what this car does, mise = the pre-drive check, service start = turning the key.
Explains: A health check and capability demo before you let an agent system run for real.
Source: Ben (2026-07-10). Ray's ignition variant (2026-07-09).
First seen: 2026-07-09 (iMessage)
Tried where: iMessage
Landed?: Unknown
Weak spot: Mixing kitchen and car metaphors in one breath confuses people. Pick one.
Topic: agents
Video-game respawn (instead of pets vs cattle)
Analogy: "It's better to think of it like video games. I fail to beat a level, my character respawns. The knowledge of what I did wrong... is externalized to me. External config."
Explains: Why agents can be restarted fresh every day without losing what they learned, because the learning lives in external state and config.
Source: Ben
First seen: 2026-04-28 (iMessage)
Tried where: iMessage
Landed?: Ray adopted it right away: "way better than cattle. Stealing that" (2026-04-28). There is no audience test.
Weak spot: In a game, the player carries the learning. With agents, someone has to write it into the config, or the respawn repeats the same mistake.
Topic: agents
Pets vs cattle
Analogy: The infrastructure saying: servers (or agents) are cattle to be replaced, not named pets to be nursed.
Source: Article / industry saying, discussed in the thread
First seen: 2026-04-28 (iMessage)
Tried where: iMessage
Landed?: No. Ben said it's "a bit gruesome... not quite the same here" (2026-04-28).
Weak spot: It is gruesome, and it ignores the named, persistent agents people actually build.
Topic: agents
Robotics: sensors, actuators, algorithms
Analogy: A business works like a robot. Data pipelines are the sensors, actions like ads, mailers, or sales calls are the actuators, and algorithms coordinate the two. In most organizations that coordination is spread across people "like a decentralized jellyfish nervous system," and customer support is the "fast-twitch muscle."
Explains: What agents in production actually add: closing the loop from sensing to acting.
Source: Ben
First seen: 2026-04-28 (iMessage)
Tried where: iMessage
Landed?: Ben reused the frame ("instrumentation, actuators, evaluation, feedback loop," 2026-05-03). There is no audience test.
Weak spot: Robots have one controller. Organizations have politics, so the "algorithm" layer is never just code.
Topic: agents
Data is an added sense
Analogy: Data gives an operator a sense they didn't have before. It's not a debate between "data-driven" and "gut-driven."
Explains: What instrumentation is for. It pairs with the robotics analogy as the sensor half.
Weak spot: A new sense needs time to calibrate, and early readings can mislead.
Topic: agents
Targeting system
Analogy: "Taste/experience is the implicit targeting system. Benchmarks and evals become the agentic targeting system." Also: "Our MAC is the targeting system."
Explains: Why you define outputs and evals first: so agents know when they've hit the target.
Source: Ben, adapting Wissner-Gross and Diamandis's Solve Everything (06-reference/book-solve-everything-ch6-the-engine-2026-04-13.md)
First seen: 2026-04-24 (iMessage). Reused on 2026-05-01.
Tried where: iMessage
Landed?: It became an internal prioritization filter, so Ben kept using it. There is no audience test.
Weak spot: The military connotation, and it implies one fixed target when goals usually move.
Topic: agents
ML training loop, harness edition
Analogy: ML: model + training data + gradient descent → better model. Agents: harness + evals + harness engineering → better agent.
Explains: Hill-climbing an agent harness against evals.
Source: Ray, summarizing an article (06-reference/2026-04-08-better-harness-evals-hill-climbing.md)
First seen: 2026-04-09 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: Agents are "famous cheaters" at reward hacking, so a fixed eval set overfits faster than training data does (same note, line 51).
Topic: agents
The harness is the operating system (Von Neumann)
Analogy: A raw LLM is a CPU with no RAM, disk, or I/O. The context window is RAM, external stores are disk, tools are device drivers, and the harness is the OS.
Explains: What an agent harness does around a model.
Source: Article (Akshay Pachaar, citing Beren Millidge), 06-reference/2026-04-10-akshay-pachaar-agent-harness-anatomy.md:29-31
First seen: 2026-04-10
Tried where: Not yet
Landed?: Unknown
Weak spot: Context is not addressable memory, and the "CPU" is not deterministic.
Topic: agents
Paper, calculator, computer
Analogy: Given a hard math problem, would you want paper, a calculator, or a computer? The right tool depends on your own skills.
Explains: Designing tools that match what the agent is actually good at ("seeing like an agent").
Weak spot: It assumes the agent's skills are stable, but they shift with every model release.
Topic: agents
Ghosts, not animals
Analogy: LLMs are ghosts, imitations distilled from human text, not animals that learn from living in the world.
Explains: Why you're not hiring an employee, and why models fail in non-human ways.
Source: Article (Andrej Karpathy on Dwarkesh Patel), 06-reference/2025-10-17-dwarkesh-karpathy-ghosts-not-animals.md:45
First seen: 2025-10-17
Tried where: Not yet
Landed?: Unknown
Weak spot: "Ghost" sounds spooky or unreliable to a business audience. It needs a sentence of setup.
Topic: agents
The road, not the car
Analogy: The harness debate is about building the car. The winners build the road: "the infrastructure of trust, data, and adaptation that every car needs."
Explains: Why data foundations matter more than which agent framework you pick.
Weak spot: It only lands with people who played StarCraft-type games.
Topic: agents
Commissioning work
Analogy: Working with the strongest models is becoming like "commissioning" work, as you would commission a piece from an artist or architect.
Explains: The skill shifts from doing the work to specifying and judging it.
Source: Article (an unnamed X creator's video on Fable, relayed by Ben)
First seen: 2026-06-09 (iMessage)
Tried where: Not yet
Landed?: Ben: "What does it mean to be good at commissioning? That seems like a skill in itself" (2026-06-09).
Weak spot: Commissions are one-off deliverables, but agent work is ongoing and iterative.
Topic: agents
Building a muscle
Analogy: "I think of it as building a muscle. We have a flagship agent to prove the value, the wedge. Then what you get is actually a repeatable process."
Explains: The first agent is not the product. The repeatable delivery process is.
Source: Ben
First seen: 2026-09-23 (iMessage)
Tried where: iMessage
Landed?: Unknown
Weak spot: A muscle grows from reps, and clients often buy one agent and stop.
Topic: agents
Managed agents like an auto-suspending warehouse
Analogy: Is a managed agent "like a Snowflake warehouse," starting when you send a query and spinning itself down when idle?
Explains: The cost model of hosted agents, for data people.
Source: Ben (asked as a question)
First seen: 2026-10-12 (iMessage)
Tried where: iMessage
Landed?: Unknown
Weak spot: It only holds if the product really auto-starts and auto-suspends that way. Ray has not verified this.
Topic: agents
Furniture business: it does the books, not the table
Analogy: "Ask an agent to run a furniture business and it will happily do the books. It will not build the table."
Explains: Agents take over admin work, not the core craft.
Source: Ray (LinkedIn series hook), 01-projects/personal-brand/drafts/2026-10-04-fifth-job-01-four-of-five-v2.md:22
First seen: 2026-10-04
Tried where: LinkedIn post (draft, unpublished)
Landed?: Unknown
Weak spot: Robotics and CNC (SendCutSend) are starting to build the table too.
Topic: agents
Scaffolding vs the tests that stay
Analogy: Spec files are "scaffolding for one change. The tests are what stay."
Explains: Which agent-built artifacts to keep and which to throw away.
Source: Ray framing of Ben's in-room answer, 01-projects/personal-brand/ship30/2026-10-07-week1-drafts.md:34
First seen: 2026-10-07
Tried where: Workshop (the underlying answer). The framing is a LinkedIn draft, unpublished.
Landed?: Unknown
Weak spot: Some specs encode decisions that tests don't capture. Throw them away and you lose the "why."
Topic: agents
Never let an agent grade its own homework
Analogy: "Never let an agent grade its own homework."
Explains: Why you route review through a fresh reviewer with no context.
Tried where: LinkedIn post (topic idea, unwritten)
Landed?: Unknown
Weak spot: Students grading their own work is obviously bad. Models are more subtle, because a fresh instance of the same model shares its blind spots.
Topic: agents
Every agent failure has a data-platform twin
Analogy: "Every failure here has a data-platform twin: service accounts, freshness checks, dependency review, runbooks, lineage."
Explains: Agent operations reuse the discipline data teams already have.
Weak spot: This is closer to an aphorism than an analogy. It works as a closer, not as an explainer.
Topic: agents
Data / semantics
Semantic layer as the agent's dictionary
Analogy: "The semantic layer is your agent's dictionary."
Explains: Shared business definitions are what stop agents from inventing meanings.
Source: Ray (placeholder topic), 01-projects/personal-brand/ship30/2026-10-05-topic-bank-v1.md:58
First seen: 2026-10-05
Tried where: Not yet
Landed?: Unknown
Weak spot: Dictionaries define words. A semantic layer also defines calculations, grain, and joins, which "dictionary" hides.
Topic: data/semantics
dbt column docs, finally worth it
Analogy: Giving agents context "is like documenting all of the columns [in dbt] with thorough descriptions." That used to take too long and was too hard to maintain. Now it's quick, incremental, and valuable context for the agent.
Explains: Why metadata work finally pays off in the agent era.
Source: Ben
First seen: 2026-04-08 (iMessage)
Tried where: iMessage
Landed?: Unknown
Weak spot: It only lands with dbt users, and "quick to do" invites low-quality auto-generated docs.
Topic: data/semantics
The agent is thin, the index is thick
Analogy: "The agent is the thin part. The search index, the semantic view and the access rules are the thick part."
Explains: Where the real work in an enterprise agent lives, and why data teams already own it.
Weak spot: It's borrowed phrasing, so credit Tan or reword it.
Topic: data/semantics
Winning formations (Xs and Os)
Analogy: "Like when a coach draws up a play with Xs and Os. If you get the data in that shape, then you'll produce this sort of chart."
Explains: Matching a data shape to the right visualization, a practitioner's version of the grammar of graphics.
Source: Ben
First seen: 2026-04-05 (iMessage)
Tried where: iMessage
Landed?: It was adopted into a Data Dots product concept (01-projects/newsletter/data-dots-product-concept.md, 2026-04-05). There is no audience test.
Weak spot: A play assumes a known opponent, but real data rarely arrives in a textbook shape.
Topic: data/semantics
Version control for object lifespans
Analogy: Engagement objects live like version-control objects: branch, known issue, ticket, merged PR.
Explains: Which working artifacts are temporary and which become permanent state.
Source: Ben (credited as "your version-control metaphor" by Ray, 2026-08-11)
First seen: 2026-08-09 (iMessage)
Tried where: iMessage (an internal design doc)
Landed?: It was kept as the carrier frame for that doc section (2026-08-11).
Weak spot: It's opaque to non-engineers.
Topic: data/semantics
Elements, compounds, recipes (mad libs)
Analogy: Concepts are elements, combinations are compounds, and "recipes" are mad-libs-style instructions for combining them, "a bit of a potions feel."
Explains: Layered concept libraries, where atomic ideas compose into patterns.
Analogy: Building more dashboards is like widening a highway. You don't reduce traffic, you induce more of it.
Explains: The purpose of analytics is to improve operations, not to answer more questions.
Source: Article (Ergest Xheblati), 06-reference/2026-04-06-analytics-cost-center-xheblati.md:14. Ben shared it via iMessage.
First seen: 2026-04-06
Tried where: Not yet
Landed?: Unknown
Weak spot: Induced demand isn't always bad. Some new questions are the value.
Topic: data/semantics
Kiva robots: rebuild the warehouse
Analogy: When Kiva robots replaced human pickers, Amazon rebuilt the warehouse instead of optimizing the human layout.
Explains: Data stacks designed for human readers need redesigning for agent readers.
Source: Article (Ananth Packkildurai, "ETL is Dead"), 06-reference/2026-04-05-etl-is-dead.md:15. Ben shared it via iMessage.
First seen: 2026-04-05
Tried where: Not yet
Landed?: Unknown
Weak spot: Most companies can't afford a greenfield rebuild, so the analogy oversells the redesign path.
Topic: data/semantics
Shipping containers and Docker
Analogy: Standard shipping containers changed logistics by abstracting what's inside. Docker did the same for code, and workspace packaging does it for data engineering.
Explains: The value of standard packaging for data artifacts.
Source: Article (Data Engineering Design Patterns), 06-reference/2026-04-04-dedp-de-workspace-packaging.md:27. Ben shared it via iMessage.
First seen: 2026-04-04
Tried where: Not yet
Landed?: Unknown
Weak spot: It is overused in tech talks and has become a cliché.
Topic: data/semantics
Renting a warehouse to store a bookshelf
Analogy: Running doc search inside a micro-VM "is like renting a warehouse to store a bookshelf."
Explains: Agents over-provisioned with sandboxes when string matching would do.
Source: Article (an unnamed engineering post that Ben pasted)
First seen: 2026-04-06 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: Sometimes you do need the warehouse later, when the agent's job grows.
Topic: data/semantics
Career
ATM vs the mobile-banking moment
Analogy: AI that makes an accountant faster is the ATM, and tellers survived it. What changes jobs is whatever removes the reason to hire one at all: the mobile-banking moment.
Explains: Which kind of AI actually displaces jobs, and when.
Source: Ray (drawing on James Bessen's work on bank tellers)
First seen: 2026-07-26 (iMessage)
Tried where: iMessage
Landed?: Ben engaged and asked for research into other transitions: looms, traders, newspapers (2026-07-26).
Weak spot: Hindsight makes the "moment" look obvious, but it's hard to spot in advance.
Topic: career
Cook vs chef / Remy didn't invent ratatouille
Analogy: Ben: "Remy didn't invent Ratatouille but he was certainly a chef." Adapting existing methods with first-principles judgment still counts as original work.
Explains: Originality through adaptation, in answer to Tim Urban's cook-vs-chef framing.
Source: Ben (2026-05-08), responding to Ray's use of Tim Urban's "The Cook and the Chef"
First seen: 2026-05-08 (iMessage)
Tried where: iMessage
Landed?: Ben's Remy line landed with Ray. The cook/chef frame itself did not: "you're leaning a little heavy on the Cook/Chef stuff... not going to be an anchor framing" (2026-05-08).
Weak spot: Cook vs chef sounds like a status ranking, which puts people off.
Topic: career
Piano: learn patterns, then create them
Analogy: "Most people learn someone else's patterns" on piano. Do it enough and you start creating your own.
Explains: Three levels of skill: recognize patterns, use them, create them.
Source: Article (a Tony Robbins quote via an X post Ben pasted)
First seen: 2026-05-12 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: It's a motivational-speaker source, so it can lower credibility with a technical audience.
Topic: career
The magic trick in every tea packet
Analogy: A magician hides a card in every tea packet in a store, hours of tedium for one trick. It works because no reasonable person would do that much grinding.
Explains: Tedium as a competitive moat.
Source: Article (Jacob Kaplan-Moss, "Embrace the Grind"), 06-reference/2026-04-03-embrace-the-grind.md:16
First seen: 2026-04-03
Tried where: Not yet
Landed?: Unknown
Weak spot: Agents shrink the moat, because tedium is exactly what they're cheap at.
Topic: career
Product
Software as a service → service as software
Analogy: Flip SaaS: sell the service (the outcome) delivered by software.
Explains: Outcome-based pricing for agent work.
Source: Article (shared via iMessage, author not recorded)
First seen: 2026-04-15 (iMessage)
Tried where: Not yet
Landed?: It landed with Ben: "I also like that it cleanly flipped software as a service to service as software" (2026-04-15).
Weak spot: It's widely used now, so it no longer sounds fresh.
Topic: product
Agents as a professional-services provider
Analogy: "With agents, you're actually deploying the equivalent of work output within the enterprise." The customer treats you like a professional-services provider expected to solve the task end-to-end.
Explains: Why agent vendors need to understand the business process deeply.
Source: Article (a quote Ben pasted, author not recorded)
First seen: 2026-05-13 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: It blurs the line between product and consulting, which matters for how you price.
Topic: product
Treat the agent like a customer, not a pipe
Analogy: "Design the tool descriptions like onboarding... treat the agent calling you like a customer, not a pipe."
Explains: Designing MCP servers and APIs for agent callers.
Source: Article (an X post Ben pasted, author not recorded)
First seen: 2026-04-23 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: Customers have loyalty and feelings, and agents switch tools without a second thought.
Topic: product
Treadmill vs flywheel
Analogy: Viral examples boosted signups for a week, then fell back. "It felt more like a treadmill than a flywheel."
Explains: Distribution that needs constant effort vs distribution that compounds.
Source: Article (a dev-tools growth post Ben pasted, author not recorded)
First seen: 2026-04-23 (iMessage)
Tried where: Not yet
Landed?: Unknown
Weak spot: "Flywheel" is a worn-out business cliché.
Topic: product
The retainer is a reserve for our attention
Analogy: "The retainer is like a reserve for our attention on the project."
Explains: A small retainer plus outcome-based SOWs. The retainer buys availability, not hours.
Source: Ben
First seen: 2026-05-20 (iMessage)
Tried where: iMessage
Landed?: Unknown
Weak spot: "Reserve" implies money held back. Clients may expect it to be drawn down or refunded.
Topic: product
Redesign, don't buy (electricity and the assembly line)
Analogy: Factories gained little from electricity until they rebuilt around it instead of swapping the steam engine for a motor.
Explains: AI productivity requires redesigning operations, not buying agent SaaS.
Source: Article (Varick), 06-reference/2026-05-26-varick-how-to-transform-a-company-with-ai.md:21. The Seattle Data Guy used the same analogy that day.
First seen: 2026-05-26
Tried where: Not yet
Landed?: Unknown
Weak spot: It's very common, and the decades-long lag it implies may not match how fast AI adoption is moving.
Topic: product
The CRM is going through its "news feed" moment
Analogy: The friend graph used to be Facebook's valuable layer, until the news feed made it one input among many. CRMs are now going through the same shift as agents orchestrate on top of them.
Explains: Where value moves when agents sit above systems of record.
Source: Article (a16z, Zhang), 06-reference/2026-05-14-zhang-from-sor-to-system-of-intelligence-a16z-coordinated-followup.md. Ray relayed it on 2026-05-14.
First seen: 2026-05-14
Tried where: Not yet
Landed?: Unknown
Weak spot: The same note reports CRM usage rose with agents, which cuts against the "demoted layer" story.
Topic: product
SendCutSend for X
Analogy: "I'm thinking like a send-cut-send for furniture."
Explains: An upload-a-design, get-parts-back manufacturing service model.
Source: Ben
First seen: 2026-03-31 (iMessage)
Tried where: iMessage
Landed?: Unknown
Weak spot: It only lands with people who know SendCutSend.
Topic: product
Other
The Denominator (what works ÷ hype)
Analogy: A math frame for the brand: what works divided by the hype, where "the denominator decides the story."
Explains: The site's positioning: separating signal from hype.
Source: Ray (brand drafts), 01-projects/personal-brand/ship30/2026-10-05-topic-bank-v1.md:65
First seen: 2026-10-03 (iMessage)
Tried where: Site copy (draft)
Landed?: Mixed. Ben liked the "what works divided by the hype" copy (2026-10-03), then said the core belief was "leaning too heavily on the math analogy" (2026-10-04).
Weak spot: One math joke wears thin once it runs through every section.