What Endures When Execution Is Free

I think about this constantly. It is existentially relevant to Hera, and I want every engineer we hire to have a sharp answer to why here, why this? Here is what I have come to believe.

The enduring work is the long tail.

Healthcare. Robotics. Self-driving. Core AI itself. These problems were not ignored for the last twenty years because they were boring; they were ignored because they were impossible. Transformers and the broader modern stack did not make them easy; they made them merely very difficult. Meanwhile, most of what was hard in 2015 (most SaaS, most mid-complexity backend, most of the classic "wedge" startup playbook) is now trivial. If your moat in 2026 is that you built a pretty interface over a database, I do not know what to tell you. The problems worth a decade of your life are the ones that were inaccessible before and are now finally cracked open.

Opinionation will not commoditize at the same rate as execution.

An LLM can write the feature. It cannot yet tell you whether the feature should exist, what bar it should hit, where to invest disproportionately, where to cut cleanly, what quality feels like when a family is using your product to care for their mother. These are judgments, and judgment compounds. Brand is the residue of thousands of those judgments made well, over years. It is slow to build and, crucially, slow to copy. This is exactly the property you want in a durable asset. Investing in taste, in quality bars, in opinion, will feel expensive right up until the moment it is the only thing your competitors cannot mimic.

Build novel data assets.

This is an old playbook and it is getting more important, not less. The open internet is about to become the cheapest, most commoditized diet an LLM can eat. The edge is proprietary, hard-won data: self-driving traces, care outcomes, what actually worked when a specific treatment was tried with a specific patient profile. Data that your business generates as a byproduct of serving people well, that compounds with every interaction, and that is valuable enough that other serious players in your ecosystem want access to it. If you are not structurally generating a novel data asset by doing your core work, your core work is probably replaceable.

Do something that matters to ordinary people.

For a long time, technology ate its own tail. We built dev tools for the people building dev tools for the people building dev tools. I don't begrudge it; I did it too. But the real opportunity of this moment is that engineering talent can finally be pointed at the problems that affect regular people's lives: aging, health, care, housing, the plagues of the everyday. There is going to be a backlash against this technology as the pace of disruption grows. There already is. The best response we have is to make sure our work is observably, specifically, humanly good for people who are not in tech. For me that is our grandparents and parents. For you it might be something else entirely.

Design toward being a net job creator.

Automation is coming for nearly everything. You can ride that wave cynically or you can design against it. At Hera, this means building a new profession — senior care experts — where the platform carries the minutiae and the human carries the humanity. Expertise and capacity become independent levers; we scale both. The cheap win is automating the humans away. The harder, better design is unbundling expertise from hour-for-hour labor, so more humans can do more meaningful work at a higher floor. There are many more industries where this shape of business is possible, and far too few people building them truthfully.

None of this is a guaranteed path. It is a bet on long-horizon problems, on taste, on proprietary data, on mattering to real people, on designing work instead of eliminating it. If that bet sounds right to you, we should talk.