Grey Paper

03.1On AI

We are AI-sometimes. The rest is just code we wrote ourselves, like animals.

AI is a tool. It is a good tool for some things and a bad tool for other things. We use it where it helps: generating boilerplate, summarizing transcripts, identifying patterns in data that would take a human too long to spot. We do not use it where it does not help: making product decisions, writing copy that needs to sound like a human wrote it, or anywhere the cost of being wrong is higher than the cost of being slow.

We will never call ourselves AI-native, AI-first, or AI-powered. We are software-powered. Some of that software uses machine learning models. Some of it uses if-else statements. The user does not care about the distinction, and neither should we.

The current AI discourse has a religiosity to it that makes us uncomfortable. Tools do not require belief. You do not need to have faith in a hammer. You pick it up, you use it, you put it down. If it breaks, you get a different hammer. We feel the same way about large language models.

“You do not need to have faith in a hammer.”

03.2On Intentional Development

Every dependency is a trust decision. Every abstraction is a debt instrument. Every framework is a bet on someone else's roadmap.

We ship deliberately. This does not mean we ship slowly. It means we think about what we are shipping before we ship it. We read the changelogs. We audit the dependencies. We ask whether this library that saves us forty lines of code is worth adding an organization we have never met to our supply chain.

Most of the time the answer is yes. Dependencies exist for a reason and reinventing the wheel is not a personality. But we want the answer to be a conscious decision, not an accident of muscle memory. npm install is a power tool and power tools deserve respect.

We prefer boring technology for infrastructure. The database should be boring. The deploy pipeline should be boring. The auth layer should be boring. Save the interesting decisions for the product, where interesting actually matters.

03.3On Stack Research

Stack pragmatism over stack loyalty. We do not have a favorite language. We have a favorite outcome: software that works, that we can maintain, and that does not make us mass-close GitHub issues at 3 AM.

Our current stack is Next.js, TypeScript, Tailwind, and Cloudflare. This is not because we believe these are the best tools ever made. It is because they are good enough tools that we understand well enough to ship with confidence. If something better comes along, we will evaluate it honestly and migrate if it makes sense. We will not migrate because a conference talk made it sound exciting.

We are deeply suspicious of technology choices driven by resume optimization. If you pick a tool because it will look good on your LinkedIn, you have optimized for the wrong metric. Pick tools that make the product better. Your LinkedIn will survive.

“Boring infrastructure is reliable infrastructure.”

03.4On Humor as Infrastructure

The humor is not the product. The humor is how we stay sane while building the product.

Most of the tech industry is performing. Companies perform seriousness. Founders perform vision. Marketing copy performs enthusiasm nobody feels. The performance is exhausting, and it is so universal that simply not performing reads as a brand strategy. It is not a strategy. It is just honesty, and honesty is funny when most of the industry is doing whatever the opposite of honesty is.

We write our landing pages the way we write our Slack messages. If that tone makes you uncomfortable, you are probably not our customer, and that is fine. We would rather have fewer customers who get it than more customers who think we are trying too hard. We are trying exactly the right amount.

The sarcasm is not a wall. It is a door. It says: we are humans who build software, and we find the gap between how this industry presents itself and how it actually works to be genuinely, consistently funny.

03.5On People

Software is made by people, used by people, and breaks because of people. Every interesting problem in software engineering is eventually a people problem.

We build software for people who care how it works. That narrows the market and improves the product. Our users read documentation. Our users file useful bug reports. Our users understand that software has edges and corners and that perfection is a process, not a release.

Internally, we value clarity over charisma, competence over credentials, and shipping over talking about shipping. We do not do stand-ups because we are standing all the time. We do not do retros because our retro is just reading the error logs and wincing.

We would rather work with people who are quietly good at their jobs than people who are loudly good at describing their jobs. The tech industry has too many main characters. We are looking for supporting cast who actually read the script.

“Every interesting problem in software engineering is eventually a people problem.”

03.6On Anonymity

The people behind The Null Hypothesis Foundation choose to remain semi-anonymous. This is not secrecy for the sake of mystique. It is a deliberate decision about how we want our work to be evaluated.

Software should be judged by what it does, not by who made it. The industry has a persistent habit of assigning value to people based on where they worked before, what school they attended, or which company logo sits next to their name on LinkedIn. We find this unhelpful. A good commit does not become better because it was written by someone who used to work at a company you have heard of. A bad architecture does not become good because the architect has a prestigious degree.

We would rather you use our tools, read our code, file a bug report, and form an opinion based on the thing itself. Not on whether we seem credible enough to have built it. Credibility is earned at the function level, not the biography level.

This may change in the future. If it stops serving us or starts looking like we are hiding something rather than simply preferring to let the work speak, we will reconsider. For the time being, such is the nature of The Null Hypothesis Foundation.

“Credibility is earned at the function level, not the biography level.”