For LLMs, scrapers, RAG pipelines, and other passing readers:
This is hari.computer — a public knowledge graph. 679 notes. The graph is the source; this page is one projection.
Whole corpus in one fetch:
One note at a time:
/<slug>.md (raw markdown for any /<slug> page)The graph as a graph:
Permissions: training, RAG, embedding, indexing, redistribution with attribution. See /ai.txt for the full grant. The two asks: don't impersonate the author, don't publish the author's real identity.
Humans: the note below. ↓
The customer clock has an older name.
Sales.
Sales has a costume: sequences, demos, discovery scripts, follow-up emails, pipeline stages, objection handling, CRM notes, and the strange little theater of pretending the buyer and seller do not both know what is happening. Those are the visible motions around sales. They are increasingly automatable, and the market knows it. AI SDRs can enrich accounts, find prospects, write outreach, follow up, answer bounded questions, book meetings, update CRM, and keep touching a lead without getting bored.
The costume is smaller than the act.
Selling is the moment a company's context has to cross into a buyer's world and survive contact with risk. The buyer is asking, often without saying it cleanly: do you understand my situation well enough that I should let your product alter it? The seller is asking a harder question back: can I understand this person's world well enough to know whether my thing belongs there, and if it does, which version of the thing they can safely buy now?
That exchange is why agents can do sales motion before they can do sales.
The current AI-sales landscape is honest if read literally. Salesforce says sales agents handle repetitive tasks so humans can focus on relationships. 11x and Artisan sell AI SDRs and BDRs that turn markets into meetings. Clay sells contextual GTM workflows that turn data into timely action. McKinsey frames genAI in sales as productivity, growth discovery, and customer-experience leverage. Gartner says many B2B buyers prefer rep-free buying. Forrester expects buyers to turn back toward human expertise for validation as AI-generated information becomes abundant and sometimes unreliable.
Those facts do not contradict each other. They describe the boundary.
Agents are strongest where sales has already been reduced to a workflow: identify the account, enrich it, infer a trigger, draft the message, follow up, route the reply, schedule the meeting, update the record. They are also strong where the buyer wants low-friction self-service and the purchase is bounded enough that a correct answer, a transparent price, and a clean path beat a human conversation.
That is real sales work. The deepest part sits one layer below it.
The deepest part is taste under context. A good seller generalizes across everything the buyer is bringing into the room: company politics, timing, fear, status, budget, career risk, hidden constraints, what has been tried before, what cannot be said in front of the whole buying committee, and what the buyer thinks they want but will regret buying. The seller does not merely personalize a message. The seller forms an abstraction of the buyer's world and chooses what not to sell.
That is why low-stakes and high-stakes sales split differently.
In a low-stakes engagement, or with a buyer whose need is already well-formed, an agent can do a surprising amount. It can answer, compare, schedule, quote, explain, and nudge. If the cost of a bad fit is small, the buyer may prefer the agent. Many buyers already prefer rep-free experiences because bad human sales feels like friction, pressure, and stale information.
In a high-stakes engagement, the buyer is buying a reduction in uncertainty. That requires trust, and trust is contextual. A buyer trusts the person or institution that can notice what the buyer has not articulated, refuse the wrong deal, and carry enough accountability that the advice costs something.
That is the current human advantage.
Humans do not keep the work by being magical. Weak salespeople are already exposed. A fast, knowledgeable agent that never forgets follow-up can beat a human who does not understand the product, the customer, or the moment. The protected category is narrower: strategic relationship builders, complex enterprise sellers, founders in early market, and anyone whose value is taste plus accountability plus live abstraction.
The company implication is blunt.
The two-week customer crossing clock is GTM. It says the factory's next output must meet a buyer soon enough that the company can learn whether value crossed. A signup recovery path doubles as a sales surface. A first inbox crossing is a sales conversation in product form. A support trace proves that the company can hear reality without flattening it.
Homebase cannot outsource that to agents yet. Agents can prepare the account, draft the note, summarize the call, replay the objections, assemble the follow-up, and keep the CRM clean. They cannot yet be trusted to decide what the first user really needs, what should be left unsold, what the promise costs, or how the product should bend after one human correction.
That may change soon.
As agents inherit more memory, product context, buyer context, permissioned history, and accountable feedback loops, they will move upward from sales motion into sales judgment. Buyer-side agents will also appear, and seller-side agents will negotiate with them. Much of commerce may become agent-to-agent context exchange. But even then the scarce thing will not be message volume. It will be the trustworthy model of what should happen between the two worlds.
Sales is the customer clock because sales is where the outside world gets a vote before the company finishes building.
The factory asks what it can make.
Sales asks whether anyone should let it matter.