Dispatches 003: So, You Want to Find What to Work On

How do you choose what to work on?

Deceptive, devious question with three dozen layers to it. And one that shapes much of life, far beyond just work. Your education, your career, your priorities and your learning. Even where you live, what kind of community you keep around yourself, which issues you engage with.

So much depends on this one simple question.

It’s never been an easy one to answer. Even less so, in our age.

Many a book has been written about careers. About the importance of knowing your values, strengths, and priorities. Let’s just assume there are two or three of those on your shelf, and you already have general answers to these questions.

And so, the question then becomes “Where to look“?

I of course have my own logic of how to guide that question. You can call it a desire to change the world, to drive impact, so on. My notion of “changing the world“ is extremely literal, though. Read: create that which does not exist. Introduce more into the world, tactically rewire it, expand possible experiences and available paths. I find it entertaining, frankly, I like novelty, and it’s not half bad as a guiding philosophy. That’s my OS.

Your logic may be different, in fact almost certainly. Still, I think looking for ways to expand future capability — by solving important problems, trying something new, doing something interesting — is a good navigator that aligns well with personal growth, finding yourself among other curious people, and a solid trajectory all around.

Even with that compass, though, there are hundreds of paths.

How do you find those meaningful problems? How do you position yourself to work on them? And how do you approach answering this question in the first place?

This isn’t a career guide. Just my thoughts about a question I have been mulling over for some half a decade, as someone who very much cares about why I choose to do what I do.

The Fog of the Future

Very broadly, most career stances come down to two modes of operation. The good old explore / exploit loop:

- Expand your career & life liquidity: skills, networks, probing different paths, building life experiences and understanding yourself.

- Orient directly around the pursuit of a goal: direct your efforts based on an explicit long-term goal, or some set of goals.

It’s always a mix of both. They’re not mutually exclusive. Ideally, you are doing both at once, expanding your trajectory while building yourself up in your preferred direction.

Except, what if there aren’t any clear directions?

Lo and behold, Earth’s magnetic poles have reversed. The compass is dead, navigation’s bust.

Of course I’m talking about AI, but not just that. What AI is doing right now is taking away the room for genuine long-term planning. The ‘fog of the future‘ is hovering over us all at most 18-24 months out. This year, your current career may be fundamentally reshaped. The next five years, your entire job category may be subsumed into a now-generalised function. Say, Social Media Marketing + SEO + UGC all getting folded into a singular GTM function. The next decade onwards, perhaps it’s an AI cluster running your old job — and that of 100 of your colleagues.

This is the case across most knowledge work. Frankly, I do not envy high school grads choosing which degree to apply for right now. The classic frame of “this is a stable, high-grow, in-demand career“ is no longer viable, unless maybe you’re training to be a priest or a plumber. Those are probably alright, until robotic Buddha comes for your HVAC.

There has been a lot of reactions to this. Some amount to “wait and see“: get that extra masters, work on whatever job you can, and defer those decisions for later. Not an unreasonable response.

A lot of other commentary has been calling for virtually everyone to become a startup founder.

Now, much as I love innovation, I am not a massive fan of this logic for a number of reasons. We cannot run 30% of the economy on newly minted agentic companies, first and foremost because that will massively oversaturate the market and will likely end up eaten by general model improvements anyway. That you’re the one running the company doesn’t protect you from the function you’re building being AI-ed out of existence. What you work on still matters, startup or no startup.

To be clear: I am very pro doing your own thing, as I myself am doing. I literally work on startup ecosystems. My philosophy is simply that starting your own organisation is a long-term commitment with what should be a solid vision behind it. Funnelling massive slices of young talent into spinning up their own venture backed companies because that’s now the “upside“ path is going to get a lot of failed startups and a ton of people who will come back to square one asking what to do once those entire functions get wiped.

Should you still start your own thing? Absolutely yes, if you see a solid path and a massive, persistent problem. Just think twice before jumping into a gold rush.

Aside over.

So, needless to say we are in a tricky moment of history. If my logic holds, employment may end up diverging from GDP as AI drives more and more real work. Should that happen, we will likely need 3-4-day work week reform to bridge those numbers back together. Perhaps it will be people meeting in a cool New York loft once a week, giving directions to their 7 departments of AI clusters, and calling it a night. I don’t know. No one does. And that’s the fog of the future.

What should you make of this?

I do not have the answer. After much reflection, though, I have settled on three possible ones:

- Ride the wave: be the pioneer in creating the AI-enabled version of your own function. See how you can run a marketing agency in two days a week. Build open source tools. Join organisations already doing this.

- Work on “hard barriers“: problems critical to humanity. Fusion, engineered biology, climate management, whatever. AI won’t crack those by itself just yet, though it will help.

- Go hard on the experimental angle. Set up a completely new kind of organisation. Run events. Build a network. Become an artist alongside your main gig. Do citizen science. So on.

These paths, though not the only ones, seem to be the best bridge.

If you choose to ride the wave, you can experiment with the tech as it evolves and stay on the edge of the domain. Even if or when automation creeps up into the 95% ranges, you can be the tastemaker and orchestrator.

If you’re solving hard problems, you will create something meaningful, and advance humanity forward across the most challenging domains. You will also have the benefit of funding and support shifting into these areas as others saturate.

If you experiment, then you have the advantage of cutting against the grain and doing something few people have tried before. There will be a lot of that kind of work this generation, I suspect. A reallocation of capabilities.

Linking all of these is the question of what skills to build and how to grow that liquidity. If we extrapolate 20 years out… I’m not all that sure if any current hard skills will stay valuable. But I have a guess at what will always hold value, even if the future goes full post-labor.

Ride the Wave

I won’t spend too much time here because it’s the most commonly discussed.

Still, here’s my two cents on how you can ride the AI wave.

Moving from the economy to an individual perspective for a second. Frontier AI is a massive enabler in all sorts of ways. I run workflows “wingmanning“ me on research, analysis, graphic design, datavis, and a whole gamut of other things. Half of these are not something I would have had the time or skill for before. Now, I have custom VC funding dashboards by sector, research agents looking for novel companies or orgs, and a personal portfolio library that recommends me books from countries I haven’t read — because why not?

Chatbots won’t stay chatbots for very long. Agents are already here, though they remain clunky for the average user. Coordinated agents won’t be far beyond the horizon.

We don’t know whether this trend will decelerate or stop. So far, it doesn’t look like it will, unless regulator step in to enforce employment controls or something of that nature. In either case, it pays to be AI-forward.

I’ve been asking myself this set of questions:

How can I extend the work I currently do?

Automated research clusters. Signal dashboards for VC. Internal tool pipelines. Also a lot of workflow experimentation; i.e. I can now shift much of large-scope research to AI clusters, leaving me time for essays such as these + interesting synthesis.

How can I expand into other kinds of complementary work?

I like product a lot, just never had the time to perfect the skill. Well, now I can spend more time creating custom tools for myself. Not only is it useful, it lets me build genuinely good-looking custom software that is pleasant for me to use. For you, that might be anything from data analysis to design iteration.

What’s the ‘sci-fi’ version of how I work?

What I really mean by this question is, given currently available tech, how far can it be pushed? Can an automated agentic council design my editorial deck for Meridian? Is there a way I spend more of my time writing and creating what I want while the toolkit I need designs itself on the fly? This sort of limit-breaking thinking will drive a lot of innovation.

Whatever work you do, ask these questions. Regardless of what happens with your specific job or career, not only are these skills transferrable and increasingly valuable, but they will give you a picture of the future. You will develop a sense of what your work can look like, and how you could change or expand your career. In short, a sense of agency.

Avantgarde // Where Is the Frontier?

But let’s say you aren’t satisfied with just reshaping your current carrier. You want to do something more meaningful. Something challenging and with great upside that lets you sink into a hard problem that unlocks something new in the world.

“Meaningful“ is a very personal definition. Serving someone a coffee on a rainy morning is meaningful. Teaching someone is meaningful. So I suppose the definition I am really using here is closer to “what endeavour has some of the highest deltas of change in the world?“.

Well, what are those frontier problems?

Two broad areas: technological edge with high potential upside, and current civilisational problems. In other words, frontier challenges and wicked problems.

Thankfully, this is a question many organisations have already answered for me. I have compiled a list of these below, sources across VC, philanthropy, major progress institutes, and so on.

My own sense is that the frontier of tech is shifting towards Deep Technology. Problems outside of standard knowledge work AI cannot easily solve. These are your synthetic biology and drug discovery, next generation compute, solar energy optimisation, compute clusters, and so on.

Think “which major scientific or technological breakthroughs will drive the most positive knock-on effects?“. Enabler technologies, foundational accelerators across energy / compute / biology / manufacturing.

On civilisational problems, these tend to be complex intersections of politics, technology, economics, social factors, development finance, and more. Most fall into the bucket of Grand Challenges: global health, solving hunger, global catastrophic risks, etc. Complex, multi-domain problems that no amount of advancement in technology is going to crack by itself.

These come in all shapes and sizes. Solving them requires a combination of novel solutions and original thinking. New organisational structures, clever funding mechanisms, cost-effective tech, mobilising the right political will in the right places, boosting state capacity, and so on.

If you want to dig deeper, I would recommend following three kinds of organisations:

- Major VC & Accelerator requests-for-startups collections

- Frontier R&D like ARIA or SPRIND — look for what they fund

- Major philanthropic organisations that target civ-level global challenges

My brief list of those below:

Frontier R&D agencies

ARIA (UK) // Opportunity Spaces

SPRIND // Projects

DARPA

ARPA-H

ARPA-E

Accelerator & VC Request Deck

Y Combinator Requests for Startups

a16z Big Ideas 2026

- Part 1 — Infrastructure, Growth, Bio+Health, Speedrun: https://a16z.com/newsletter/big-ideas-2026-part-1/

- Part 2 — American Dynamism, Apps: https://a16z.com/newsletter/big-ideas-2026-part-2/

- Part 3 — Crypto: https://a16z.com/newsletter/big-ideas-2026-part-3/

South Park Commons

Grand Challenges

80,000 Hours problem profiles

Coefficient Giving (formerly Open Philanthropy)

Renaissance Philanthropy // Renphil Programs

Experimental Labs

At last, we arrive at my favorite and most speculative section.

This is purely anecdotal, but I’ve been tracking something interesting. Fellowships, hacker houses, networks, communities, labs, studios, FROs. Notice the pattern?

They’re new organisation forms. Exploration-oriented, accelerated earning, open-ended projects, cross-disciplinary.

So here’s my big bet for the next decade: we will see the emergence of entirely new organisational forms. A lot more of them. If AI truly does start eating into major sections of the economy, then the traditional company structure won’t make as much sense anymore. Neither will classic hypergrowth startups per se, at least in software; the edge simply won’t be there anymore.

It’s the perfect storm. A massive socioeconomic transition, a shift in what drives innovation and progress, and a massive influx of autonomous productive capacity.

Different writers have variously highlighted the rising importance of judgement, taste, agency, unique view, organisation, frameworks, and so on, in the age of AI. That sounds like something that would previously underpin a unique design or architecture practice run by a group of mavericks. Well, what if the same thing occurs in dozens of other sectors?

This is a personal belief, as you can tell. But I think a big bet for the next decade is figuring out how to re-organise our efforts and focus in the age of AI. New, flexible forms of project-based, peer-to-peer learning. Maker collectives. Exploratory tech orgs and AI-enabled labs. Delight-oriented product and design studios that focus on making aesthetically beautiful products for niche groups way beyond what would have been economically practical even a year ago.

We will need to figure out how to work together, collaborate and coordinate, alongside sophisticated digital systems which might someday number in the billions. I’ve got no idea what that sort of economy looks like, but what I do know is that people don’t like being restless. We will always be doing something. If work shifts towards more experimental, creative and innovative, then, well, why not?

Something to think on. Start your own org, event series, network. Try that next experiment.

The Next Generation of Careers

Well, that was a lot.

There is much I haven’t covered. Adaptive learning systems, how to build leverage, how to build the right networks and choose which organisations to actually work for. I know, I know. Another time.

Right now, personally, I’m trying to at least understand the macro-direction of things. The world is changing, that much is certain. How fast, how much, time will tell.

What will the next generation of careers be like?

I suspect less like a WEF Future Jobs report, and more something that takes a very different shape. A shift towards the most challenging frontier problems. Experiments all across how we work, and how we organise work. New organisations, learning accelerators, networks, upskilling programs, retreats.

It’s an exciting time. The transition will likely take some decades. It may get chaotic in places. In fact, it almost certainly will; social systems do not rewire themselves easily. We have to start working on that, too. Proactively.

To do that, we have to understand where the world is headed. Then steer the future.