Yes! It has actually made me even stronger at thinking how to imagine the solution since I know if I explain it well the agent will execute well. Being in CS this was always a skill we learned but shrinking problems and defining them is now how the world must operate exceptionally now.
Writing the code was always partly a proxy for understanding it. When the proxy gets free, the request and test has to harden.
Also it requires upfront decision making (that use to be easy to slip past when talking to humans) now will go off on a tangent that immediately affects your competitive speed and cost.
So, machines forcing us to be better is the moral of the story.
Here he mentions "Store trajectories " is a tent pole to see vectors of work over time which goes back to the point above. But also, will that trajectory be stored in the agent/framweork, or will it also need live elsewhere for us humans?
like with most agentic dev I do these days, I go between "I need this" to "I used to need this when humans were involved but is it just vestigial" a lot. In this case, why am I documenting at all if the agent is pretty good at understanding things quickly via the context and indexes it creates from the code itself.
...on the other hand... since we still have humans using the features and interacting with them, knowing what is going on and why it made a decisions (for better or worst) doesn't seem like something to let go of.
What a cool concept to turn the SM 140 chars concept into a programming demo list! Is there a reverse to this that work to compress a JS script into 140 chars to help show some creative ideas without needing to fully understand the code? Also, how good is AI at generating these in your experiences?
A friend of mine is using N8N a lot, but he isn't a developer and I think that is the main reason. I think LangChain is the most popular, but there is always a trade off between upstart costs, and the cost of forcing a pre-fab to fit your needs as you grow. I personally have used NextJS's AI SDK a a lot since it is very web friendly (nodejs/JS) https://ai-sdk.dev/docs/introduction.
Interesting. I use a lot of python and nodejs for my apps, maybe this shell angle is a weird quirk I developed for no particular reason. I keep a seperate repo called “manager” and use it to manage my other parts of the stack like, “run tests for each repo, bump version, commit to git” with a single CLI command. I have no idea why that started. My guess is an agent created a shell script and I just went with it and built up around it. Good to hear other perspectives. Thanks!
There is a reshaping of what we think is “right”. Like, I am less and less interested in SDKs since I can just ask for a lightweight, custom, integration into anything I want based on API docs. I mean, I still use a lot of SDKs…buuuut it feels like I’m forcing the AI to use them for my monkey brain’s sake. Is predefined mashup of structured services even necessary anymore?
Dude - crazy. When I saw this post I was like - finally someone said it. But it isn't just "iPhone". Why is spell check so bad EVERYWHERE..... still?. Like, how is it I am still even able to share texts, emails, etc that have mistakes at all? I feel like "spell check" is so old school. Intent, and matching intent, without typos is way over due. A bit meta: but, I am going to have to re-read this post - why? I still send texts that say "What re you doing?" - hwy?
Finance people are funny. They are so wrong when you hear their logic and references, but I also realized it doesn't matter. It is trends they try to predict, fuzzy directional signals, not facts of the moment.
Writing the code was always partly a proxy for understanding it. When the proxy gets free, the request and test has to harden.
Also it requires upfront decision making (that use to be easy to slip past when talking to humans) now will go off on a tangent that immediately affects your competitive speed and cost.
So, machines forcing us to be better is the moral of the story.
reply