dgunay

8 ms·

91.
by dgunay 1y ago
What about any tool, language, library, or codebase that is unnecessarily complex? Should we never bother to put in the effort to learn to use them? It doesn't mean they are without value to us as programmers. For better or worse, the
92.
by dgunay 1y ago
I haven't seen anyone talking about using agents in this way. I wonder if it would be helpful for learning e.g. a new language or codebase to have the human write the code while the agent takes the role of the "backseat driver&quo
93.
by dgunay 1y ago
I don't see the problem. I don't necessarily think it is good for society, but IMO producing erotica is nowhere near as much of an "AI safety" concern as fomenting violence, distributing misinformation, or facilitating p
94.
by dgunay 1y ago
For anyone who either doesn't want a GUI-based workflow or wants a lot of custom behavior, I would encourage you to roll your own tool, either by yourself or with an agent. I did this at work and the tool became very usable in short or
95.
by dgunay 1y ago
It's a different kind of fun for me. I've been enjoying seeing my agents produce code while I am otherwise too busy to program, or seeing refined prompts & context engineering get better results. The boring kinds of programmin
96.
by dgunay 1y ago
In CS you can leave the buy zone immediately. I don't necessarily believe that Valorant's decision to fence players in their spawn for the first minute during the buy period is simply to save on server costs, especially because th
97.
by dgunay 1y ago
In addition to the other arguments presented here, I have 24/7 location sharing enabled with my wife so that neither of us has to actively remember to check our phones all the time. And if it's potentially an emergency, it is stil
98.
by dgunay 1y ago
I've found that it can be hard or expensive for the agent to do "big but simple" refactors in some cases. For example, I recently tasked one with updating all our old APIs to accept a more strongly typed user ID instead of a
99.
by dgunay 1y ago
I use it as more of a focusing tool. Without it, I frequently get distracted by small rabbit holes (I should add more logging to this function, oh I should also add some doc comments, oh I should also refactor this, etc) or I don't hav
100.
by dgunay 1y ago
Most of the AI generated code I review is pretty much okay. Usually does what it should and meets some standard of quality. But it usually looks and feels just slightly stylistically foreign compared to the code around it. I personally edit
101.
by dgunay 1y ago
My uneducated guess would be that rent and labor are much cheaper (in relative terms) in Japan than in the US, perhaps so extremely that it dominates compared to the marginal cost of producing food.
102.
by dgunay 1y ago
The economy depends on some level of growth, so if we can't accomplish that with a stable or shrinking population then it's gonna be a bad time for a while. EDIT: I did not think I'd have to state this explicitly, but: yes, I
103.
by dgunay 1y ago
We had the technology to push out a vaccine in less than a year. Modern medicine is of course smarter than it was a century ago. What went poorly is our society's collective response. From the medical and governmental establishment, th
104.
by dgunay 1y ago
I see a lot of debate about the definition and merits of unit, integration, and e2e testing here. For my workplace, we have recognized a few issues with an overreliance on unit tests. When you have important behaviors and invariants enforce
105.
by dgunay 1y ago
I have been doing this, and it does sort of work, but the problem is that for things that can't easily be turned into deterministic lints, prompting isn't 100% reliable. Every bit you go against the LLM's training data, it&#x
106.
by dgunay 1y ago
Letting go of the particulars of the generated code is proving difficult for me. I hand edit most of the code my agents produce for taste even if it is correct, but I feel that in the long term that's not the optimal use of my time in
107.
by dgunay 1y ago
When some of those plates are unnecessary, it causes problems. I may be responsible for 3 plates worth of work, and be capable of 6 on a really good day. On a bad day though, 3 may be all I can handle and an interruption=dropped plate.
108.
by dgunay 1y ago
Having driven in both LA and NYC+NJ, LA drivers feel almost serenely calm to me most of the time. In LA, as long as you don't do anything obviously stupid and give plenty of room for people to see you coming, people will just chill and
109.
by dgunay 1y ago
I don't personally care whether bikes (or scooters) ride on the road or the sidewalk, but my one ask is that: If they ride on the sidewalk, they should behave like pedestrians. That is, do not blast into the crosswalk at 20mph (impossi
110.
by dgunay 1y ago
Claude may be arguably the best model, but why decide unilaterally for your users that they _have_ to use it? If there's no particular feature that only Claude offers, this is just needless vendor lock-in. And what happens if another l
111.
by dgunay 1y ago
Here's my personal workflow. I've built a relatively small (<5k LoC Go) CLI to facilitate it. Mostly vibe coded in a dogfooded fashion. When I want to start an agent, I run a small command that takes in a prompt (and some optio
112.
by dgunay 1y ago
Codex-cli does use MacOS sandboxing by default. It does unfortunately cause issues for my workflow because the agent is very restricted in what it is allowed to do (like, read/write the Go build cache) and its command whitelisting conf
113.
by dgunay 1y ago
I know it's due to safety but I really wish there were an economy version of the lie flat seat. I'd gladly sleep in a set of bunk beds stacked 3-4 high for an 8 hour flight.
114.
by dgunay 1y ago
I just wish the AGENTS.md standard wasn't a single file. I have a lot of smaller context documents that aren't applicable to every task, so I like to throw them into a folder (.ai/ or .agents/) and then selectively cat t
115.
by dgunay 1y ago
I like them because they're easier to launch multiple instances of and take fewer resources. Being able to fire agents off into tmux sessions to tackle small-fry issues that they can usually oneshot is a powerful tool to fight the deca
116.
by dgunay 1y ago
Played with it a bit. So far it lacks some key functionality for my use case: I need to be able to launch an interactive session with a prefilled prompt. I like to spawn tmux sessions running an agent with a prompt in a single command, and
117.
by dgunay 1y ago
We did a lot of this where I work to make using cucumber easier, and when we later stripped cucumber out it still was incredibly useful.
118.
by dgunay 1y ago
I've commented elsewhere on lack of buy-in from stakeholders as a huge pitfall. On a technical level, cucumber is also at odds with the need for a test suite to be easily maintainable. What I mean is that each test (especially e2e test
119.
by dgunay 1y ago
Yeah this was unfortunately my experience too. A lot of head nodding and a token effort up front but they quickly reverted back to playing telephone on Jira tickets. A year or two later we're slowly ripping out the cucumber tests and r
120.
by dgunay 1y ago
For Copilot the quality of the tab complete is less of an issue for me than the fact that it is often very slow, or doesn't trigger at all when I would expect it to. I'll sit there feeling like an idiot for 10 seconds and then gla

More ›