Just moved my whole setup from @bot to open source @mausbot_ , running on my own subscriptions.
I expected to lose polish. Instead I got more: more features, better sync, and nothing deciding for me what I can plug in. What changes is how much of the app is actually yours.
The closed ones win the first hour. Open source wins the second month.
9
Everyone's learning AI from the top down. Interfaces first, fundamentals later. That order is why so many demos work and so few products do.
Two years into models that can plan, call tools and run for twenty minutes straight, the real skill isn't prompting. It's building the thing around the model.
Four areas that will matter more than any framework you pick up this year:
1) Agents.
Break the job into steps, keep context small, decide when the model plans and when it just executes. The hard part isn't the API call. It's step nine, when the plan is wrong and nobody noticed.
2) Harness.
The loop around the model: how tools get called, what goes into memory, what gets dropped, what the agent can touch, and how evals prove it works instead of a good feeling. Most agent failures live here, not in the model.
3) Workflows.
Turning a clever agent into something repeatable. Same input, same output, every time, without someone watching. That's where the money is, and where a lot of them stop.
4) Systems.
Databases, caches, queues, retries, idempotency. Unglamorous and decisive. Agents break exactly here: two of them writing the same row, a job running twice, a queue backing up while a model thinks.
None of this is exotic. Agents are software, a harness is a runtime, evals are tests, workflows are pipelines. The only new thing is that the component making the decisions is a language model, and it's confidently wrong sometimes.
Learn them as one stack, not four separate hobbies. Then when something breaks, you know which layer broke. That alone saves weeks.
Pick the layer you're weakest at and go deep on it this month.
1
7
Most people don't have a motivation problem. They have a starting-line problem.
"Work on the landing page" has no first step, so your brain files it under later. "Write the headline, three versions, ugly is fine" does.
Before you close the laptop tonight, rewrite tomorrow's first task until you could start it in 30 seconds.
Then tomorrow you're not deciding anything, just continuing.
9
Claude has Haiku, Sonnet and Opus. Google's building a periodic table of models now: Hydrogen, Argon, Carbon. What a comeback.
introducing Gemini 4 Hydrogen
this new model brings cutting-edge intelligence to everyone — with speed and efficiency for everyday coding and agentic workflows
start building with Hydrogen via the Gemini API and in AI Studio: aistudio.google.com/prompts/…
7
Before an agent writes any code, ask it one thing: "what will you NOT touch?"
The list it gives you is the scope. If a file you care about isn't on it, you just found out where the surprise diff was going to come from.
When it starts "cleaning up" something off the list, you already have the sentence to point at.
Try it on the next task you hand over this week.
16
You just spent 3 fucking months watching courses and another 2 months building projects.
You finally finish one, and a week later you've already got another "million dollar idea" in your head.
So you start building again.
And again.
And suddenly you're working on 5 different projects at once because now you know how to delegate to 18 AI agents and run everything in parallel lol.
But here's the problem.
You've gotten really good at building shit, but you have absolutely no idea how to get people to actually care about it.
No traffic, no distribution, no audience, no clue how to sell the value of what you've built.
And instead of learning any of that, you just start another project.
Because building feels productive. Watching another course feels productive. Setting up another agent feels productive.
But none of that means you're building a business.
Meanwhile, some random guy builds a simple product in one day and spends the next 7 days getting it in front of people, talking to potential customers and actually selling it.
And guess what? He's probably making more money than you.
Not because his product is better.
But because he understands something you don't.
Building the product is only half the game. Getting people to want it is the other half.
And most of us are so obsessed with the first half that we never even play the second.
26






