Loved that puzzle, thanks for sharing it. I’ve solved a lot of math problems in the past but this one had a unique flavor of interleaving logical reasoning, partial information and a little bit of arithmetic.
Defensibility re: companies is well studied, by Hamilton Helmer who and his grad student did case studies of several hundred "defensible" companies and identified 7 "powers" that help companies defend against specific competitors. Note that each power is only applicable towards a specific competitor, and not as a blanket statement. The powers are 1. economies of scale, 2. network effects, 3. counter positioning, 4. Switching costs, 5. Branding, 6. Cornered resource and 7. process power.
About: Pioneer's mission is to coordinate the funding for rapid decarbonization. We take the pain out of the government grants application process by using dozens of LLM workflows to reduce the effort required to apply and win government awards, and we’ve helped companies win $160M to date. We have more demand than we can serve, and we’re growing revenue.
Culture: Our core values are kindness, impact, intentionality, initiative, feedback and efficiency in that order, and we implement these throughout our processes.
You: Enjoy the 0-to-1, hungry for growth, actively collaborate and pair with users, can handle the nondeterminism of LLMs, start sentences with “What if…“, enjoy building across the full stack.
Stack: We use TypeScript, React, Next.JS, LangSmith and other modern tools.
Yeah! That's the danger with any kind of "model" whether it is CoT, CrewAI, or other ways to outsmart it. It is betting that a programmer/operator can break a large tasks up in a better way than an LLM can keep attention (assuming it can fit the info in the context window).
ChatGPT's o1 model could make a lot of those programming techniques less effective, but they may still be around as they are more manageable, and constrained.
This is so uncannily close to the problems we're encountering at Pioneer, trying to make human+LLM workflows in high stakes / high complexity situations.
Humans are so smart and do so many decisions and calculations on the subconscious/implicit level and take a lot of mental shortcuts, so that as we try to automate this by following exactly what the process is, we bring a lot of the implicit thinking out on the surface, and that slows everything down. So we've had to be creative about how we build LLM workflows.
Maybe one way to think of it is fractal management where a manager would have deep read-write interactions with few skip levels. HBR style says read everywhere, write only direct reports. And it makes for a good software design except that humans are not computers, but there is a shared global context - the company vision and mission.
Through fractal management, a visionary leader can have a better chance to ensure that the vision is translated into practice at the various levels of detail.
Fractal management is only part of it, though as it is a technique, but it doesn’t cover the enormous skin in the game founders have about the success of the company. For many founders, the company is their baby(I am projecting here) and they want to make it succeed. Contrastingly, many of the professional fakers instead see it as just a job, and a step on the ladder. Principal/agent. Without genuine care, and cohesive vision, fractal management can quickly devolve into chaos. It is high reward and also higher risk!!! Maybe that’s why only the founders do it but not their VPs. I wonder if any VPs at Airbnb are doing anything remotely similar to what Bryan Chesky is doing as management style? (Honestly I have no idea)
I am sure that many founders failed also because of it as they might have been missing the charisma, clarity and conviction to pull this off.
(PS. Take my ideas with a big serving of salt, I am a founder but not at a large organization, and the article mainly focuses on large orgs)
Tony Fadell’s idea of a Parent CEO vs Babysitter CEO in Build comes close to this founded mode concept but I am so glad PG memed the founder mode concept into existence, and talk about the gaslighting. And I bet he is right about there are many founders operating somewhat that way already but without a shared mental model and a name to it.
It is really hard to put in practice because of all the gravitational pull towards mediocrity.
About: Pioneer's mission is to coordinate the funding for rapid decarbonization. We take the pain out of the government application process by using LLMs to gradually reduce the effort required to identify, qualify, apply, and comply with government awards. We are passionate about climate impact and creating a supportive growth environment based on the fundamentals of Radical Candor. We also have a proven business model and rapidly growing revenue.
Culture: Our core values are kindness, feedback, intentionality, ownership, and impact. We implement these in various creative ways: toms, laps instead of sprints, feedback instead of 1:1, and decision journals! Our newest team member said “Every interaction is “kind”, coming from a good place of respecting each other.”
You: Enjoy the 0-to-1, pre-PMF phase, have an owner mindset, collaborative, have curiosity for technology like LLMs, start sentences with “What if…“, enjoy building across the full stack.
Stack: We use TypeScript, React, Next.JS, and other modern tools.
P.S. We are also hiring for a Founding Operations Lead - apply through the same link as above.
Most recently I wrote a series called “how to be a game changer” which focused on creating win win situations in all aspects of life. One of the essays “Opt out of cynicism” made it to #2 on HN.
I’ve been slow to write for the last 1.5y as I was giving a birth to a startup.
About: Pioneer takes the pain out of the government application process by using LLMs to gradually reduce the effort required to identify, qualify, apply, and comply with government awards. We are passionate about climate impact and creating a supportive growth environment based on the fundamentals of Radical Candor. We also have a proven business model and rapidly growing revenue.
Culture: Ask us about our culture: toms, laps, feedback, and decision journals! Our newest team member said “Every interaction is “kind”, coming from a good place of respecting each other.”
You: Owner mindset, collaborative, have curiosity for technology like LLMs, start sentences with “What if…“, enjoy building full stack, and enjoy the 0-to-1 phase.
Stack: We TypeScript, React, Next.JS, and other modern tools.
It is not that the commit was wrong. It was the right commit, in the sense that it did expose and trigger the bug. OP is being too modest :)
But the root cause of the bug was in another earlier commit. In a complex system such as the linux kernel, there could be multiple contributing factors. Both the "introducing a sleep" and the "root bug" are contributing factors, but we as engineers tend to think of the "root bug" as the actual problem, because we like more elegant explanations. Also, the "root bug" is likely something that goes against the ethos of the software project, and "common sense best practices", to a much larger degree than "introducing a sleep" is.
Another example of how real world data is always messier than a neat model:
Vatican is entirely encircled in Rome, therefore any area outside Vatican is closer to Rome than the Vatican, yet if we look at them as point sources, as this map does, you see a lot of other area closer to Vatican.
I am reminded of one of Matt Mochary’s teaching: when you have to do a task, which takes away your energy, and cannot be delegated to someone else, them make it “exquisite”.
Next time I have to “pick up trash”, I will try to think “what would these samurai do?”
This. Most of the value of the company is to be created in the future, and you'll likely pivot (based on statistics, your specific idea may turn out to be spot on).