Rob and Marc discuss AI-powered startups, the lack of spoons at the DoD, and why you should never take your eyes off the worm.
Rob is working on a DoD contract building out services for space launch services, trying to tackle Elon Musk’s lament that he can build rockets faster than he can get access to the range to launch them. Universal Documentation System (UDS) governs this onboarding regime, and Rob’s team is tackling digitizing it to be a simpler way for managing this. They crushed it, moving fast and re-using pre-accredited and hardened code to get value for the public money. Learned the magic of transparency and humility with the SMEs and operators. Government stakeholders super excited, and they celebrated with random ice cream of unknown origin.
Then Rob went back home to Savannah where Hurricane Helene had knocked out power unexpectedly, and was creeped out by how unprepared most folks are for mild inconveniences. If we can’t survive this, how can we make it through anything? Good fun family time for the apocalypse. At least we had plenty of Nuggets to build forts.
Marc bought a potentially life-changing device for less than $60. An automated cat feeder with a day and night vision camera, a two-way microphone and speaker, and remote feed release that is now outside in the chicken coop. The possibilities are endless, at least until it gets rained on and breaks. Worth it.
When you start obsessing about a problem, ask yourself why you are letting people you don’t even respect live “rent free” in your head. Marc was wondering how to solve this problem. Then I realized, wait, do I really need to solve this problem at all? It might actually be impossible. But like the kid in the matrix said, you can’t bend the spoon. What you need to realize is there is no spoon. In many cases, there is no real problem at all. It’s just you and your thoughts. You must bend yourself.
Rob asks “Why is this everywhere?” Should everyone just join a small startup based on trust and empathy between a few people? People feel more fluid and confident in what value they’re producing and what they’re responsible for. Rob wonders why more people don’t do this? Marc points out that great 4-person startups don’t grow on trees. Maybe that’s the startup idea – helping people find each other and join to create these little trustworthy startups. Rob mentions he’s been working on ways for 4 people to get together with high trust and isolate themselves from the poison of larger organizations.
The government is so bad around this, basically tons of massive jobs programs where people get paid outrageous amounts for little to no usable output. The types of things they’re doing could be done onshore with proven professionals for $25k. Yet the DoD pays millions for partnerships with fake entities that never deliver results. Rob is looking at a new way of working and running an experiment with micro contracts facilitated by LLMs, where larger contracts are interpreted into several 1099 roles that can be broken out and made accountable.
Marc has a question around LLMs now that he’s got a subscription to ChatGPT. Why did the simple task of generating a little man riding a sandworm from Dune (requested for reasons left unexplored) go so poorly. The first attempt was in the ballpark, but the sandworm had big eyes and teeth like a dragon, unlike real sandworms on Dune. The next prompt was for a small tweak – to remove the eyes and teeth from the worm. ChatGPT said “no problem” and delivered a new image – it was a similar picture, but the worm’s mouth was closed (almost hiding the teeth), and the eyes got bigger instead of being removed. WTF?
But why can’t AI take the eyes off the worm?
Bottom line is you have to own it and finish it yourself. AI will likely only get you to 80%. The last 20% is all you. Never forget that you have a great assistant now, but never delegate the end results. AI can make you rough drafts, and give you ten variations on something, but you must choose the best of the ten.
Rob calls out people who use AI as “word butlers”. He tried to get it to be imperfect and it couldn’t be weird and quirky like a human. As cool as automation and synthetic relationships are, it’s the opposite of why we exist and do what we do. Marc mentions how Douglas Rushkoff uses AI for drafting the next chapter of what he’s been writing (after secretly already writing it). If it matches what he wrote, then he throws it out for being too predictable.
Its critical people are on the same page for what problem they are trying to solve. More important is whether it is the right problem to solve. Teams can knock something out of the park, but if it’s not in the 20% of the critical work, they will still fail. Like ChatGPT and the worm.
Are you solving the problems that the world needs you to solve?