Content
@
https://opensea.io/collection/dev-21
0 reply
0 recast
2 reactions
Darryl Yeo đ ïž
@darrylyeo
My two cents: using AI tools âcorrectlyâ necessitates understanding the fundamentals. You can probably get by on AI prompts alone if your team values speed of execution over quality, but assuming everyone else is doing the same just understand youâre still easily outpaced by someone with a âmuscle memoryâ of identifying patterns and syntax unique to that programming language / tech stack. By all means, use AI tools to bridge your gaps and accelerate learning (I do this all the time now for parts of the stack Iâm unfamiliar with). But if Iâm hiring you to work on a specific problem area, at minimum Iâd expect you to have read lots of relevant docs/articles/videos, be extremely comfortable testing/evaluating different libraries/tools, make opinionated decisions and execute on them, which is easier the more experienced you are.
3 replies
2 recasts
13 reactions
Phil Cockfield
@pjc
We've had "technical debt" for time immemorial; that's not a new thing. It's the phenomenon of doing something quickly, trading off downstream maintainability or scaleability, for some other more expedient reward. Remaining entirely "tech debt" free is a sub-optimal stance, as those expedient rewards (often "discovering something real" in user-space, or grabbing customers quick on a new idea etc)...is a thing you manage very carefully, and refactoring as fast and strategically as you can afford, to avoid the inevitable complexity explosion and the project implosion that is the sword of Damocles forever hanging overhead. Hickey laid it out (in a different context from AI gen tools) in his famous talk "Simple Made Easy" (2011), but when I think about your comments @darrylyeo, it occurs we're talking about the same damn thing. Gen AI code slop gets you to project complexity implosion faster. A tradeoff that's often smart, but should be seen as such. ref: https://www.youtube.com/watch?v=SxdOUGdseq4
1 reply
0 recast
2 reactions
Darryl Yeo đ ïž
@darrylyeo
Really good points! I think generative AI excels at prototyping new features and saving keystrokes over syntactic details, but keeping codebases reliable and sustainably human-maintainable will very much continue to be a subjective and âhuman-drivenâ task. The tradeoffs engineers have to consider very much depend on the stage of the project, the ecosystem itâs a part of, the design goals it optimizes for, and how closely the act of transforming it into a desirable or âgood enoughâ state maps to your pre-existing knowledge and experience of solving similar problems.
1 reply
0 recast
3 reactions
