Is Our Brain Enough for This World?

I've lived in Italy, Berlin and San Francisco. Lately, I've been travelling: London, Slovenia, Berlin again, Cyprus, Delhi and Bangalore in India, and Shenzhen. I love meeting people, and moving around like this means I get to do a lot of it.

It also means I forget things. I can remember that a conversation was interesting without being able to explain what made it interesting. Or remember a person but struggle to place them. Was that Berlin or London? Did a friend introduce us? What were we going to do next?

Some of these people become friends. Others might become colleagues or help with something I'm building. Either way, I want to remember what they told me. It's frustrating when I can't, especially when I know I cared about it at the time.

Being a founder adds to this. You're trying to build a company while meeting the people who might help you do it. Each conversation gives you something else to think about, often before you've had time to finish thinking about the last one.

Who can help us with this?

A company needs money, customers and people who can do work the founding team can't do alone. Finding them takes time. You have to explain what you're building, hear why someone isn't convinced and work out whether they're right. Often, you need a conversation just to understand what you should be asking for.

This gets called networking, but that word doesn't tell you much about the work. A customer explaining why they won't buy can be more useful than someone saying they love your idea. An engineer might point out a problem you'd otherwise spend a month discovering yourself.

In a diary study of 94 CEOs of major Italian companies, Oriana Bandiera and colleagues found that 85% of the time in their detailed records was spent with other people. These were established-company CEOs, not startup founders, and the records covered activities lasting at least 15 minutes. Still, that's a lot of a working week spent in conversation. What Do CEOs Do?

Of course, you can fill a week with meetings and get very little done. Y Combinator's Aaron Harris recommends concentrating fundraising into a period of time rather than letting investor meetings drag on alongside everything else. You still need time to build the company you're telling people about. Aaron Harris on fundraising

Even when you choose the meetings carefully, you have to do something with what you learn in them.

The ten minutes between calls

Imagine a morning with a call every half hour: a potential customer at nine, an investor at nine-thirty and a candidate at ten. Twenty minutes each. On paper, there should be time between them.

Then the first call runs over. A teammate needs an answer. You haven't looked at the next person's profile yet. The notes can wait until lunch.

Suppose the customer needs approval from a colleague who wasn't in the meeting. The investor offers an introduction once your prototype is ready. The candidate wants to meet the team before deciding whether to continue.

By evening, you remember three promising conversations. You might even remember the investor's offer clearly, except for the part about waiting for the prototype. So you send a message asking for the introduction, and they have to explain the condition again.

You didn't forget the whole meeting. You forgot enough to misunderstand what was agreed.

There is also the problem of what you heard in the first place. In this example, you might spend the start of the investor call thinking about the customer's objection. You're listening, but you're also trying to solve something else.

Researcher Sophie Leroy calls this attention residue: some attention stays on the previous task after you switch. Her research shows how unfinished work and interruptions can leave less attention available for the next task. Leroy on attention residue

Being quick to understand things doesn't give you your full attention back. There may be a detail you never properly took in, however sure you are that you were listening.

This pressure extends well beyond founders. Microsoft's 2025 Work Trend Index surveyed 31,000 knowledge workers across 31 markets. It reported that 48% of employees and 52% of leaders described their work as chaotic and fragmented. The survey didn't measure memory loss, but it does suggest that a lot of people recognise this kind of working day. Microsoft's report and methodology

I don't read that as proof that our brains can't cope with the modern world. It does make me question the expectation that we should move through days like this and keep track of everything unaided.

I know you. Where did we meet?

Recognising someone and remembering your conversation with them are different things. A face can feel familiar while the details of the encounter are hard to retrieve. Researchers studying familiarity and source memory examine this distinction: recognising something doesn't necessarily tell you where you encountered it. Cash, Papesh and Harrison on face recognition

For the kinds of conversations I'm talking about, those details are often what you need most. A contact entry tells you someone's name and company. It might not tell you that a friend introduced you because you'd both run into the same manufacturing problem.

And even if you remember the conversation, you can still forget to act on it. Remembering to do something later is called prospective memory. You might recall exactly which document you promised to send when someone asks about it on Friday, having completely forgotten to send it on Tuesday.

That's an uncomfortable mistake because the other person can't see what happened. They don't know whether you forgot, got busy or changed your mind. A small lapse can leave them wondering whether to follow up at all.

You don't know what you'll need later

Writing down the next action helps, but some conversations don't have one.

Suppose someone mentions a manufacturer they've worked with. You have no reason to ask about it then. A month later, you run into a production problem and realise they might know the answer. Now you're trying to remember who mentioned it, possibly across several cities and dozens of conversations.

Or a friend says they're thinking of leaving their job. Weeks later, someone asks whether you know a person with their experience. You already have the phone number. What you need to remember is that they might be open to the introduction.

You can't predict all of that while you're listening. A useful detail can seem unremarkable when you first hear it.

And plenty of it has nothing to do with work. If a friend tells me about a difficult situation at home, I want to remember to ask how it's going. I don't want to have the same conversation from the beginning because I've forgotten something they already trusted me with.

Notes worth coming back to

Before adding AI to this, there are a few ordinary habits worth taking seriously.

After a meeting, write enough that you could understand the note next week. Include how you know the person, what mattered to them and what you agreed to do. "Good call. Follow up" won't help much. "Wants to meet the engineering team before continuing" gives you something to act on.

Keep any conditions attached to the promise. In the investor example, "introduction after the prototype is ready" would have prevented the misunderstanding.

If something needs to happen on a particular day, set a reminder. Otherwise, you've solved the problem of storing the information but still have to remember to look for it.

Sam Gilbert studied reminders for delayed intentions in four online experiments with 1,196 participants. Reminders improved performance on the tasks, and people used them more when memory demands or the likelihood of distraction increased. This was research on experimental tasks, not founders or AI, but it gives evidence for the usefulness of external reminders. Gilbert's study

It may also help to write down where you stopped before an interruption takes you elsewhere. In four studies, Leroy and Theresa Glomb found that brief plans for resuming work reduced attention residue when people anticipated time pressure on their return. A practical version could be: "Checked the supplier's quote; next, confirm whether shipping is included." Leroy and Glomb, 2018

These notes can live in whatever you already use. There's no need to reorganise your whole life before trying to leave yourself a better record of one meeting.

What I'd want a software brain to do

The limitation is that you still have to write the note, remember where you put it and find it again. That's where I'd like more help.

Before meeting that candidate a second time, I'd like to ask what we discussed and get a short answer: the technical problem, their question about the team and the introduction I still owe them. Then I should be able to open the original note or transcript and check it.

That's what I mean by a software brain. It could use notes, messages and conversations captured with permission to help me find details I would otherwise have to search for myself. It would also need to handle the less obvious question: who was the person who told me about that manufacturer?

I'd want it to be careful. "I might be able to help" cannot turn into "they agreed to help." If two people share a name, their conversations can't get mixed together. And when there isn't enough information, I need to know that, rather than get a convincing guess.

This is part of why I'm building Draki. I keep coming back to conversations that happen away from a screen, where something useful can be said and never written down. I'd like to be able to return to those details without making everyone repeat themselves.

You can use the Draki app on its own. For me, the bracelet is about keeping capture conveniently available as I move through the day, and being able to turn it on or off whenever I want. The current app sends recordings through Draki to OpenAI over Wi-Fi after permission, and uses cloud text processing for summaries and search. I can set up Draki's custom MCP connection in an AI assistant that supports remote MCP, depending on my account and its settings. If I authorize this optional, read-only connection, the assistant gets text and context; raw audio is not shared through that MCP connection. Audio still goes to OpenAI for Draki's own transcription.

That doesn't require keeping every conversation forever. I've written separately about what's worth keeping in memory and why a larger context window isn't the same as memory. A system like this needs to let people choose what to keep, correct mistakes and remove information they no longer want stored.

I also wouldn't expect it to fix a calendar with no room to think. Fewer meetings and better notes may help more than another tool. Having a record doesn't mean I paid attention, and a summary doesn't tell me everything someone meant.

I like how many people I get to meet. I don't want to give that up. I'd like to spend less time trying to reconstruct our last conversation, so that when we speak again, we can continue it.

Sources

• Oriana Bandiera, Luigi Guiso, Andrea Prat and Raffaella Sadun. What Do CEOs Do?, working paper, 2011. Diaries from 94 CEOs of major Italian companies; not a survey of startup founders.
• Y Combinator. Aaron Harris on Fundraising and Meeting with Investors. Practitioner guidance on concentrating fundraising and protecting time for company building.
• Sophie Leroy. Research on attention residue, University of Washington Bothell. The researcher's explanation and links to the original studies.
• Microsoft. Breaking Down the Infinite Workday, 2025. The survey figures above describe reported work experience; they do not measure memory or founder performance.
• Daniella K. Cash, Megan H. Papesh and Alan T. Harrison. False Memories of Familiar Faces: The Role of Familiarity in Recognition and Source Memory, Experimental Psychology. Experiments on face familiarity and the context associated with faces.
• Sam J. Gilbert. Strategic Offloading of Delayed Intentions into the External Environment, Quarterly Journal of Experimental Psychology, 2015. Four experiments involving 1,196 participants.
• Sophie Leroy and Theresa M. Glomb. Tasks Interrupted: research on ready-to-resume plans, Organization Science, 2018.