@zanlib0

product engineer (js, elixir, react)

 Silesia, Poland
zanlib.dev
Joined July 2019
  • Tweets3.6K
  • Following57
  • Followers95
523Media
new poast, not gonna tell you anything about it because the title speaks for itself YOU STILL READ THE CODE ANON? zanlib.dev/blog/do-you-still…

"Do You Still Read the Code?" • zanlib

Avoiding interaction with the code is not progress.

zanlib.dev
1
1
218
people need to chill
1
7
i suspect the most common failure mode in fully vibecoded software systems is going to be the inability to distinguish between a bug and an overlooked requirement
2
1
6
412
interesting that claudish forced active voice seeps into code as well (e.g. a generated api seems to be more likely to have a key like "covers" instead of "coverage")
1
30
this is an absolutely fantastic movie overshadowed by the absolutely abysmal netflix series that released alongside it i seriously wish my international friends get to experience it in english, it's REALLY good
1
2
58
it does look a bit like just another period drama but what is really apropos current affairs is the portrayal of the relationship of technical/engineering/scientific entrepreneurial nouveau riche and the entrenched elites the romantic plot is almost an implementation detail
1
2
26
imdb.com/title/tt37082105/

Lalka (2026) ⭐ 7.7 | Drama, Romance

2h 42m

imdb.com
6
(above from zanlib.dev/blog/on-ai-in-the…) making those decisions transparent and discoverable in the age where thousands of microdecisions of how to write your code, what to name each variable, &c. are automated brings us back to the CI/CD case again

On AI in the Current Year • zanlib

zanlib.dev
1
10
the teams that stay consistent will win, the teams that amplify the randomness from LLMs will have issues
3
i think it comes down to designing harnesses for the multiplayer codex and claude code are generally local software, designed to be used by one person. how do you make sure that TEAMS can work with agents effectively, without imposing undue constraints on individuals?
2
38
"When automated deployment pipelines (CI/CD) became popular, organisations with broken deployment processes didn’t get better deployments. They were either in no position to begin implementing CI/CD because their process was chaotic, or they automated broken deployment processes"
1
11
agents don't have that feel because their context windows are ephemeral, so we have to make it concrete by logging the decisions more
1
5
agents' entire "frame of reference" is ephemeral, it's context that comes up and goes away, humans in turn have much longer memories and time horizons, we have to ultimately maintain the resulting code
1
5
a human can be held responsible for a decision, an agent can't
1
7
expanding on this a bit: in this "era" of software dev, first we relied on models getting better, then harnesses getting better. harnesses are pretty good now, but designed for singleplayer. to make them work in teams xitter.cf/zanlib0/status/2110689…
zanlib@zanlib0Oct 15
one of the more annoying things about the current ai era is that it's very difficult to respond to support requests with full certainty (without spending time to dig through slop), there's always a chance that the agent is lying to you and you lie to the customer
1
35
more replies
on their own there are a ton of business decisions we'd make uncosciously when we coded features by hand, but by coding by hand we had a certain feel for the codebase and would sometimes remember when the relevant business context was brought up later
1
4
a compiler has to be usable in any domain, in any business case, or rather its abstraction boundary should run in a way to make decisions about business cases easier for the user; agents are different in that they might take on some of this business context and make decisions
1
5
but the more interesting thing is scope - comparing agents to compilers is incorrect because compilers exist apart from your specific domain they are written to be a technical layer of abstraction, with boundaries designed in a way to skirt around any specific business case
1
5
(to bring back the retarded comparison) lalamos are different from compilers determinism is the obvious case that people always cite - compilers, generally, produce the same output for the same input agents, generally, dont
1
4
we cannot rely on the fact that every single aspect of developed code was consciously decided these days, because it might have randomly come out from the agent that didn't consider all aspects of it, and didn't have all product context
1
3
anyway there maggie mentioned that she figures that preserving decision history is the big important thing now, and it's something they're working on at @GitHubNext, building a multiplayer harness that treats decisions as first class citizens
1
5
(i started thinking about this listening to this podcast btw, it's very good @GergelyOrosz @Mappletons) newsletter.pragmaticengineer…

Design Engineering with Maggie Appleton

Maggie Appleton on what engineers can learn from designers, working with AI agents, and why human judgment still matters.

newsletter.pragmaticengineer.com
1
10
it was important even before ai but the speed at which new code is being written now puts a lot more pressure on tracking what was decided consciously and what was accidental
1
6
Load more