A small team can spend months staring at a familiar problem: there’s more work to do than there are people to do it. Hiring might help, but it costs money and time, and sometimes the work itself is still too uncertain to know exactly who to hire.
Mirendil, a San Francisco AI research lab founded in early 2026 by former Anthropic and Google researchers Behnam Neyshabur and Harsh Mehta, is betting that the answer may not always be a bigger team. On November 12, the company announced a $1 billion Series A led by Kleiner Perkins and Clearlake Capital, with participation from several other investors. That followed a $200 million seed round in June, which valued the company at $1 billion.
The funding numbers are extraordinary. But for founders and small teams, the more useful part of the story is the operating idea behind them: when agents take on hands-on work, team size may stop being the main limit on what a team can attempt.
That doesn’t mean a handful of people can instantly do everything a large organization does. It does suggest a question worth testing in your own business: Which parts of the work are taking up skilled human attention, even though they could be delegated to a well-scoped process or tool?
What Mirendil says it has built
Mirendil’s tools are designed to make frontier AI research more accessible to small teams. Its agent platform, Lor, is in private beta. The company says it can run hundreds of agents that write code and carry out experiments for days at a time. Its model Fern supports a context window of up to 10 million tokens.
The company’s most eye-catching example is a coding model it calls Mirendil-Base. Mirendil says one researcher, working with agents, pretrained the model from scratch in about three and a half months, using less than $10 million in compute. The researcher then post-trained it into Fern.
Those are company-stated results, and the costs and timeline describe a specialized AI research effort, not an ordinary small-business project. But the broader point is easy to recognize: some work that once required a larger group may now be possible with a smaller team coordinating more delegated activity.
Mirendil’s progression is also striking. It announced a $200 million seed round in June, then a $1 billion Series A roughly five months later. The scale of that financing is not a playbook most founders can follow. The operating model, however, offers a more transferable lesson.
Start by increasing what one person can do
When a team feels stretched, the instinct is often to hire. That may be the right move. But recruiting is not the only way to add capacity, and it is usually not the fastest or cheapest way to find out whether a new way of working could help.
Mirendil’s story starts with changing what one researcher could do, rather than with a hiring plan for a hundred people. In a small business, the equivalent might be asking whether one person can get from customer feedback to a useful product update faster, or whether a small marketing team can produce and review more campaign variations without lowering quality.
The goal isn’t to automate everything. It’s to identify a specific, repeatable task that consumes time and test whether a tool or process can take some of the load. Start with one workflow: gathering research, sorting customer questions, drafting a first version, running routine checks, or preparing a report. Keep the test narrow enough that someone on the team can judge the result.
A useful question is: If this worked, what would one person be able to finish that they can’t finish today? That’s more practical than asking whether AI can “transform” the whole business. It gives you a defined task, a clear comparison, and a way to learn without committing to a major reorganization.
Make the demonstration before the pitch
There’s another lesson in the timing of Mirendil’s funding. The headline round came after the company had shared its initial results and examples of what its tools could do. The funding is not proof that every business should adopt the same approach, or that a large round will follow any successful experiment. It does underline a valuable principle: a working demonstration can make an ambitious idea easier to understand.
For founders seeking investors, customers, or internal support, a clear example of the new way of working is more persuasive than a promise about what it might do someday. Show the before and after. What used to take several steps? What can one person now complete? Where did a human need to make a decision? What still required correction?
You don’t need a billion-dollar lab to make that case. You need a focused experiment with an observable outcome. A short pilot that saves a few hours each week, reduces missed handoffs, or lets your team test an idea sooner can give you evidence for the next decision. If the experiment doesn’t help, that is useful information too.
When execution gets cheaper, judgment matters more
Delegating more of the hands-on work doesn’t eliminate the need for people. It changes where their attention matters most. Someone still has to decide which problem is important, how success should be measured, and whether the result is good enough to use.
For small teams, that makes judgment a core operating skill. It means choosing a customer problem worth solving instead of chasing every shiny tool. It means setting a clear brief, checking important outputs, and knowing when to stop an experiment that isn’t earning its keep.
A workflow that can produce many drafts, tests, or analyses also creates more choices. More output is not automatically more progress. If the team can generate ten versions quickly, someone still needs to recognize the one that fits the brand, serves the customer, and is worth refining.
A practical way to build that judgment is to set simple guardrails before you delegate. Decide what the tool may draft or organize, what needs human review, and what should never be sent or published without approval. Then review results against a real standard: accuracy, usefulness, customer experience, time saved, or whatever matters for that task.
The useful takeaway for founders
Mirendil’s $1 billion Series A is a headline about capital, but the more useful signal for founders is a change in what a small team might be able to attempt. The transferable part isn’t the funding, the research infrastructure, or the claim that team size no longer matters in every setting. It’s the willingness to test whether a person can do more by changing how the work gets done.
Try this as a small-team exercise:
- Choose one bottleneck. Pick a task that is frequent, time-consuming, and possible to evaluate.
- Define the finish line. Agree on what a useful result looks like before you try a new tool.
- Run a small test. Keep the scope limited, and make sure a person can review the work.
- Compare the outcome. Look at quality and time saved, not simply how much output was produced.
- Decide what deserves to scale. Continue only if the new process genuinely helps your team or your customers.
A billion dollars is not available to most founders, and it doesn’t need to be. The valuable experiment may cost little: one workflow, one focused trial, one honest review. For small businesses, capacity can still come from hiring. But it may also come from giving the people you already have a better way to use their time.
The teams that explore that possibility early may find room to build with less friction, not because people stop mattering, but because their effort can go further.
