The New Shape of a Company.
My GitHub profile picture is of me as a sophomore in college. I have a Computer Engineering degree and two years of professional software engineering out of school, but I’ve spent most of my career since then working in every other part of a business that is not directly writing code.
Last month I built the core of a company-wide “brain layer” for our customer data. Federation, ontology, provenance, per-claim verification, self-evolving context. The kind of system that Y Combinator and Alex Liberman’s LinkedIn post both describe as what somebody should build. I built a working version of this and I documented exactly how in a separate piece on proof-driven requirements.
A year ago that would have been absurd. The fact that it isn’t absurd anymore is the story I want to tell.
The old thesis was right about its time
For most of tech history, execution was the moat. It was the moat because building things was expensive in knowledge-worker hours, and those hours were the rate limit on everything. You had a vision. You could articulate it. Fine. Could you ship it? The question was load-bearing, and most of the time the answer was no.
YC’s old definition of a “technical founder” was someone who can build a full-stack app, release it, and maintain it alone. The gap between “I have a vision” and “I have a thing in production” was enormous, and closing it required a person who lived in the code.
What changed
AI collapsed the gap.
The new bar for “technical founder” is not “can you build Dropbox alone.” It’s: can you stand up a working proof-of-concept that services actual customers, get the first few users on it, retain them, and experience the momentum of more wanting it?
The bar was always can you go from vision to revenue? The team size was a function of what the tools couldn’t do, not the work itself. Then the tools had the biggest leap in their history.
What actually closes the gap
It is tempting to read “AI collapsed the gap” as “AI does the work now.” It doesn’t.
Look at how K12 curriculum gets built. Traditional textbook publishers rely on teams of hundreds — authors, editors, subject-matter experts, reviewers — to produce a single textbook every three to seven years. So what happens when the cost to produce one drops from $5 million and 3 years to a fraction of that, with both quality and alignment going up?
But AI cannot do that alone. Give an LLM a series of prompts and ask it to author a standards-aligned course and you get something confidently mediocre, because an LLM is a statistical mirror. Trained to predict the most likely next word, its output flattens unique quirks into consensus. This is what’s called “AI slop.” The people who can actually collapse that timeline are craftspeople: the ones who spent decades perfecting the craft by hand at an elite level, but who are also operating at the forefront of building with AI, and therefore know exactly what to internal tooling to build and how to buidl it. The craft comes first; the tools are there to amplify it.
This is not a thought experiment for me: building products and curriculum is the craft I have spent more than a decade in, working with Curriculum & Instruction leaders at school districts from New York to Hawaii. Today I lead the Content team at ChalkTalk, where collapsing that timeline is a project I am spearheading.
, Replit’s CEO, frames Replit’s edge this way: he spent more than a decade building coding infrastructure before AI coding took off, and he credits that head start for the lead they hold even as the foundation-model labs ship competing tools. Resarch-backed curriculum authoring, its effective delivery methods, and our instructional models are our head start.
And the payoff reaches the customer. Ryan Patenaude describes the moment well: districts today are auditing every vendor on real usage, outcomes, and cost per student, then asking the partners they trust a single question — “can you do more, so we can turn three vendor relationships into one?” A company shipping faster and better earns the right to be the answer.
The sleeper edge: internal tools
When the bar to build external products drops, the external product commoditizes. If anyone with a vision and agents can ship what used to require a team, the market floods with products, so having a good product stops being enough of a moat for SaaS companies. The moat expands from a company’s customer-facing product to also include its distribution channels and internal ops.
Most companies have treated internal tools as overhead, a thing the ops team cobbles together and the distinguished engineers don’t want to work on. But that mindset is one of the most expensive mistakes a leadership team could make, because internal tools are where AI-enabled execution compounds the most in the new shape of a company.
The company that has 3 AI-powered layers: (a) an brain layer for customer data, (b) a pipeline-intelligence layer for prospects, (c) a production layer that builds and delivers the work itself, updated from the same source of truth, outcompetes the company with the slightly-better external product. Their reps walk into every call knowing more. Their renewals arrive armed with narrative, not just health scores. Their product and published content ship the right things ahead of schedule.
If you’re tinkering with AI and looking for an efficacy test, there’s only one you should use. If your AI work doesn’t change how a colleague or a customer actually works the moment it ships, it isn’t an AI strategy. It’s just expensive bookkeeping. The internal tools that compound are the ones where the brain reaches the hand, and the bottom line is impacted.
The new shape of a company
Put it together. A company in 2026 is: (1) a vision, (2) a lean team of highly-leveraged agentic system builders, (3) a customer-facing edge that stays human because humans are still and will always be the point, and (4) a compounding stack of internal tools that absorbs bigger and bigger portions of the work every quarter.
This is also why the old way of sizing a company is about to lie to you. Funding round and team size were never execution capacity; they were a proxy for humans, how many you could hire and how fast. When someone with a vision and a thorough agentic loop builds what a fifteen-person team used to, the proxy collapses. Output per resource becomes the key leading metric, and the companies that look impressive in 2027 will post startling ARR per FTE, because they rebuilt their shape around what AI enables.
A few things follow:
- If you are a deeply curious person with a vision, whether you run a company or a corner of one, and you’ve told yourself you can’t build because you don’t code: that limitation has expired. Hold the vision cleanly and drive a fleet of agents to execute it. An engineer can help you review what you made when demand for what you shipped explodes, but you are no longer blocked by that at the ideation stage.
- If you are an engineer or a “technical founder” by the old definition: you can now do 10x more. The ceiling just moved. Whether you raise it by building harder or by lighting the older, bigger companies on fire is up to you.
- If you are an investor: round size, team size, and hiring velocity are about to betray you. The founders who found the new shape first are raising a quarter of what you’re writing checks for and out-executing the bigger-check teams. By the time the old proxies correct, the companies that won with them will already be compounding.
And remember: if your AI bet is shallow, it’s an internal cost-optimization play that gets absorbed by something smarter in a couple of years. If your AI bet is deep, it doesn’t augment your company. It reveals what your company actually is.
The biggest differentiation that matters for companies is vision. We just finally live in a world where it isn’t bottlenecked by anything else.
I’m Mohannad; I’m the Founder, President, and Chairman of the Board of ChalkTalk—a K-12 edtech company in the Curriculum & Instruction space. We recently brought in a phenomenal CEO to run our company’s growth, enabling me to go all-in on building our company’s AI infrastructure across all domains. I spend most of my days shipping that infrastructure, which reshaped how I think about what companies are for and what today’s org charts should look like. If any of this resonated, or if you see things differently, I’d love to hear it.
