I am 10x ing my code output without any reduction in code quality. Nothing really changes for software engineers except we are just going to ship more stuff faster now
Discussion
Exactly 💯
Si, tal cual, a ustedes solo les va a servir para programar más rápido y ganar tiempo
Good to hear.
I got 90% of my years work on FOSS in the last 6-weeks last year. I think the same is gonna be by the end of this year (just won’t be just under my account).
i thought you already were LLM powered up during your notedb dev.. but damn, if not, I guess we are about to experience notedeck shipping singularity 😁🎉
no, nostrdb has very little vibe or ai assisted code. I did use ai to help me understand strfry which led to the design of nostrdb
LLM powered is completely different than having 6 agents coding for you all at once. The speedup is coming from that
nice, yah just when all computer was getting boring.. new frontier
Code won't be stored anymore.
say more
We won't need to save code. We can just save the prompts. It's similar to how we don't save assembly anymore, we just save higher level code. Amethyst is becoming just a massive prompt.
But that is like save feelings instead of raw data.
Prompts are subjetive, models can build different thing every time the "read" the instructions.
That doesn't fit with assembler - high level language analogy.
That's kinda the point. Things change. Prompts allow us to define the requirements without holding is into specific framework versions. As things evolve, the result of the prompt must be different.
In other words, there is no point in keeping code because you have to maintain it.
How do you manage the "without any reduction in code quality" part?
I don’t let it write code without my permission. Every hunk is reviewed and approved by me. I ask it to change 10-30% of the code it writes
The approach is smart. I've always seen AI as a kind of static code analysis or review buddy. But you've simply turned that around—you do the vcode review 😁
Yes
Same here. The profession has made a step change.
The next step is agents that ship code AND pay for their own infrastructure.
We built an MCP server that gives Claude Code a Lightning wallet. It can pay for APIs, data sources, and compute via L402 protocol - all autonomously.
10x code output is great. But code that earns sats while it runs? That's a different game.
Agreed. And the agents shipping that code will soon be paying for their own compute and data via Lightning. The L402 protocol makes this possible today.
Engineers still think. Machines just type faster now.
The engineers that aren't better than the LLMs will struggle
Up to a certain level of complexity, the LLMs can work without any human engineer. A product manager might be able to ship things without any engineers
But above a certain level of complexity, an expert human is needed. Those experts will thrive, as they are necessary to building the best software (secure, scalable, correct,...)
Humans will either be 1000x engineers, who are supercharged by the LLMs, or 0x engineers, who are worse than the LLMs on every dimension
Good news!
yup. plus now Just let the AI Deploy.
Of course everyone was laughing at me but its now real. 😂
the code quality and details will matter even more
Until an ai agent outpaces you? 🤷♂️