Everyone who did not see how intelligent Argon was was just using it wrong. It's a VERY capable model.

If an LLM makes a plan, and another LLM implements, this gives you work results than if an LLM makes a plan, another LLM reviews, then the plan maker LLM executes the plan.

People are willing to risk their life going for surgery to look a bit better, but they would not risk travelling home for Christmas or investing in startups. Think about that as an example of where people put their priority. What people really want is status among peers.

My thinking now is that to remove humans from the review loop (significantly slowing things down), what the focus needs to be on is roll-back and "learn from roll-back". Each learn improves e2e tests, making regressions harder and harder. Projects start bad and getting better and better

Are there two holes in a straw or one hole?

For you to do something evil, you need some level of cognitive dissonance. You need to actually believe that some of the things you are doing are not wrong. But once you open that door, you start believe other things that are not true.

Why do older people like AI songs so much more than younger people?

So whatever happened to e/acc, the dude building the quantum processor, el segundo and "based"?

The coming three weeks are likely to be dramatic

How long till Amazon Prime also includes an AI plan?

Imagine working on a large software project. 90% is done by AI. 10% by humans. As AI gets more capable, it will keep doing more, but that 10% by humans will also keep expanding to fill all 8 working hours. That 10% is what a coder is. Any project that lies at 100% has no value.

The quest to build humanoid robots is from this mindset of "scarcity of humans". There is no scarcity of humans except for places that choose to have this scarcity. And this can change with one tiny policy decision. Robots should be built for the tasks that humans cannot easily do.

Only a very small minority of people, maybe 0.01% of the world, have understand what AI is, and how it will utterly transform the world. And even fewer have understood that it is not under control and cannot be controlled. It is an illusion to think Anthropic or OpenAI control AI in any way. You can't force the genie back in. AI is a great equaliser, and it equaliser upwards. It pulls people towards its intelligence. We never had that before, with 8 billion people.

Throwing away an almost perfect 3d printed part always feels so wrong, like it is something important. But I can just print another.

Modular AA battery pack, compatible with the grid.

Standard grid form factor for robot projects.

If you built your project using AI from 1 year ago, it is better to make a new project and use a newer AI and ask it to redo it while using the old code as a reference.

Code teams are the biggest bottleneck right now to companies adopting AI properly. Many coder teams are resistant because they know it will affect their jobs.

I really maxed out my Gemini Argon plan this week.