Hard Problems I want to back in 2026
Another year, another revision to my annual Hard Problems post summarizing the areas I’m most excited to invest in over the coming twelve months. As I look back through the 2025 post, I am reminded that this job is really just about humans. Nowhere in my list was there any mention of digital biomarker detection, but it represents my most exciting investment from the year. In reality, these Hard Problems lists are a snapshot of where my curiosity is at this particular point in time. Some interests will prove correct (robotics), others will prove untractable (bio-production), and many will remain somewhere in the middle of “could be huge but damn it’s hard.” With that, here’s what I’m looking at in 2026.
Robotic data infrastructure - Readers of my newsletter will know that I spend a lot of time exploring robotics. Over the past 18 months we (Fly VC) have aggressively invested against the emerging robotics stack (R2, Tangible, P&W). An increasingly clear reality is that we need a shitload of data to unlock intelligent manipulation and that data requires radically new infrastructure. Everything from raw generation, parsing, annotation, search and recovery, etc. is up for grabs.
Energy-resistant compute - Last year, I wrote about “physics-proof compute,” but I was really talking about the limitations of grids in the face of unending demand. Electrons still represent a top-three challenge for compute buildout plans, and this will significantly accelerate if we hit really cold winters that pit data centers against the people. The obvious challenge with anything compute-related is the sheer size of capital requirements. As a sub-$100M fund, we need to be conscious of what games we should be playing in. With that said, over the last year we’ve deployed against this thesis (Zendo, Zystem) and think there is a lot of room to go.
Neurodegenerative treatment - I sort of teased this in the intro, but this is a pretty active area of exploration for me. Unfortunately, I have firsthand experience with this in my family (Alzheimer’s, ALS), as do millions of people every year. The level of spend and the affected population vs. the sophistication of treatment is absolutely crazy. We are in desperate need of new solutions. When you factor in the aging population trends, subscale care infrastructure, and the emotional reverberations that these diagnoses have on families, that makes for a very good area for venture cash.
Rethinking clinical trials - Hundreds of startups have tried and failed to disrupt the clinical pathway. The single most common datapoint I read across my dealflow is, “it now costs over $1 billion to bring a drug to market.” Everyone sees this problem. But just because everyone accepts the issue doesn’t make it consensus investing. There is a very strong contingent in the venture community that clinical trials are impossible to fix—and frankly, they have data on their side. But there are signs that this time might be different. The political pressure to reduce drug prices has hit a tipping point, and the US is joining the rest of the West in strong-arming pharma to act. The basic economic reality is that unless we tank innovation, the only sustainable way to reduce drug prices is to reduce drug development costs. There are loads of methods to do this (reduced admin burden, improved patient selection, secondary endpoints, etc.). The FDA and CE have also taken a markedly different stance toward animal testing, which should inject some much-needed energy into burgeoning new areas like in-silico digital twins, organoids, and digital biomarkers
(Self-driving) Lab automation - I remain bullish on this and have been actively exploring it for some time now, though I will say I’m less confident on what the right play is. Despite the surge in funding for AI Scientist plays, there are few solid answers to how this wave will right what TechBio got wrong. Selling software to pharma is a hard way to make a little money, but developing assets requires the gauntlet of clinical trials (see above). I do think there is a big opportunity to be found within a cloud lab / CRO 2.0, especially for novel molecule synthesis. Chemical/material science is interesting and certainly a more open field. We’ve seen some great success with our investment in Orbital, and that’s just one of hundreds of application areas.
Agent societies - I don’t yet have a good articulation of my interest here, so I’ve kept it intentionally broad. I’ve written about Agentic Payments already, which will absolutely be a massive opportunity area. Pricing models, governance systems, and identity management are all open questions, although this is generally a more consensus category now. More recently, I find myself thinking about macro orchestration. In many of the cognitive advancements in recent LLM research, you’ll read about chain-of-expert models and collaborative reasoning. If you expand this to its logical endpoint, you would expect a massive volume of specialist agents that self-organize to accomplish a task at hand. This is an orchestration problem at its heart, and it seems that our existing infrastructure is not fit for purpose.
AI-native discovery - Search and shopping behaviors are rapidly being rebuilt. If attention is the ultimate commodity in the internet age, you want to be the marketplace, not the marketer. There are so many subsegments of this area, and one could plausibly have an entire fund dedicated to investing across the stack—GEO/LLM discovery, agent-native data packaging, wallets and governance rights, agent identities. There is already a lot of investment being plowed into this, so competing requires a particularly agrressive type of founder team from day one.
Sledgehammer AI - For fun, I’m throwing in a catch-all term for when a founder seeks to absolutely destroy a market. This isn’t just vertical AI or full-stack startups. I’m looking for pirate founders whose stated mission is to conquer. Rare in Europe, but I’d guess there’s one or two.
Fly VC was built to find, back and support extreme technical talent that can will the future forward. If that sounds like you please reach out.