Dev Wells
@wlls.dev
(╯°□°)╯︵ ┻━┻
Principal Eng @ Capital Rx -- focused on frontend
Principal Eng @ Capital Rx -- focused on frontend
For me it has gone downhill. It used to be very useful (maybe 3 months ago) and wouldn't veer off topic. Now I feel like I get a bunch of article summaries that aren't coherent to my original query.
It also seems to over-emphasize clickbait as opposed to actual research, which is annoying.
It also seems to over-emphasize clickbait as opposed to actual research, which is annoying.
December 11, 2025 at 2:21 PM
For me it has gone downhill. It used to be very useful (maybe 3 months ago) and wouldn't veer off topic. Now I feel like I get a bunch of article summaries that aren't coherent to my original query.
It also seems to over-emphasize clickbait as opposed to actual research, which is annoying.
It also seems to over-emphasize clickbait as opposed to actual research, which is annoying.
FWIW this also seems to be the case with languages. I've had a surprising amount of success with anything statically typed that has a small surface area (Gleam has hit this sweet spot for me lately).
December 8, 2025 at 2:57 PM
FWIW this also seems to be the case with languages. I've had a surprising amount of success with anything statically typed that has a small surface area (Gleam has hit this sweet spot for me lately).
question: how does one learn atproto? i have apps i want to build in it but not sure where to start (initial search mostly brings up more academic resources)
December 4, 2025 at 11:10 AM
question: how does one learn atproto? i have apps i want to build in it but not sure where to start (initial search mostly brings up more academic resources)
i was surprised to find that claude is also capable of spinning up lightweight hand-rolled versions of things as well (given you provide it some requirements)
i made a custom privacy-focused analytics platform for high interaction SPAs in like a weekend:
github.com/devdumpling/...
i made a custom privacy-focused analytics platform for high interaction SPAs in like a weekend:
github.com/devdumpling/...
GitHub - devdumpling/beacon: 🗼Lightweight, privacy-first analytics platform
🗼Lightweight, privacy-first analytics platform. Contribute to devdumpling/beacon development by creating an account on GitHub.
github.com
December 3, 2025 at 12:18 PM
i was surprised to find that claude is also capable of spinning up lightweight hand-rolled versions of things as well (given you provide it some requirements)
i made a custom privacy-focused analytics platform for high interaction SPAs in like a weekend:
github.com/devdumpling/...
i made a custom privacy-focused analytics platform for high interaction SPAs in like a weekend:
github.com/devdumpling/...
Interesting, why? How do you differentiate between user journeys for new users who are looking to get information vs login immediately?
I feel like this is typically what landing pages are meant to address? I agree you shouldn't land on a landing if you already have a session.
I feel like this is typically what landing pages are meant to address? I agree you shouldn't land on a landing if you already have a session.
December 1, 2025 at 2:38 PM
Interesting, why? How do you differentiate between user journeys for new users who are looking to get information vs login immediately?
I feel like this is typically what landing pages are meant to address? I agree you shouldn't land on a landing if you already have a session.
I feel like this is typically what landing pages are meant to address? I agree you shouldn't land on a landing if you already have a session.
sometimes it's absolutely ridiculous--like why are you doing this? just use the predefined script in the manifest??
but at the same time it's interesting
but at the same time it's interesting
November 30, 2025 at 8:25 PM
sometimes it's absolutely ridiculous--like why are you doing this? just use the predefined script in the manifest??
but at the same time it's interesting
but at the same time it's interesting
My goal is to open source / make the tool easy to distribute so others can run their own experiments if they want. Will shoot you the repo when I have it ready.
And yeah I'll hit you up!
And yeah I'll hit you up!
November 30, 2025 at 8:24 PM
My goal is to open source / make the tool easy to distribute so others can run their own experiments if they want. Will shoot you the repo when I have it ready.
And yeah I'll hit you up!
And yeah I'll hit you up!
I almost never do this, but not bc i'm so scared of security / privacy. I like to read the commands personally bc I find it fascinating how LLMs will use bash commands to circumvent their limitations (e.g. seen some very creative uses of awk/sed piping that LLMs taught me)
November 30, 2025 at 8:23 PM
I almost never do this, but not bc i'm so scared of security / privacy. I like to read the commands personally bc I find it fascinating how LLMs will use bash commands to circumvent their limitations (e.g. seen some very creative uses of awk/sed piping that LLMs taught me)
recently read a good book around this topic!
"Think Again" by Adam Grant
a little pop-psychy for some, but I enjoyed how it tackles exactly what you're talking about in detail and how to more intentionally revalidate and test your own beliefs (in a non-judgmental way)
"Think Again" by Adam Grant
a little pop-psychy for some, but I enjoyed how it tackles exactly what you're talking about in detail and how to more intentionally revalidate and test your own beliefs (in a non-judgmental way)
November 30, 2025 at 5:00 PM
recently read a good book around this topic!
"Think Again" by Adam Grant
a little pop-psychy for some, but I enjoyed how it tackles exactly what you're talking about in detail and how to more intentionally revalidate and test your own beliefs (in a non-judgmental way)
"Think Again" by Adam Grant
a little pop-psychy for some, but I enjoyed how it tackles exactly what you're talking about in detail and how to more intentionally revalidate and test your own beliefs (in a non-judgmental way)
My general hypotheses are:
- I ship more code with AI assistance
- I procrastinate less with AI
- Code/system quality is lower with AI (but passable)
- I don't develop as strong of mental representations with AI use (retainment)
- Moderate AI use is overall more fulfilling than heavy or none.
- I ship more code with AI assistance
- I procrastinate less with AI
- Code/system quality is lower with AI (but passable)
- I don't develop as strong of mental representations with AI use (retainment)
- Moderate AI use is overall more fulfilling than heavy or none.
November 30, 2025 at 11:44 AM
My general hypotheses are:
- I ship more code with AI assistance
- I procrastinate less with AI
- Code/system quality is lower with AI (but passable)
- I don't develop as strong of mental representations with AI use (retainment)
- Moderate AI use is overall more fulfilling than heavy or none.
- I ship more code with AI assistance
- I procrastinate less with AI
- Code/system quality is lower with AI (but passable)
- I don't develop as strong of mental representations with AI use (retainment)
- Moderate AI use is overall more fulfilling than heavy or none.
but to "rethink"/clarify my own perspective with data gathered from my actual work.
I'll write a little tool that I can use to monitor my own metrics for some weeks with no AI tools (period), moderate AI (e.g. inline complete, asking questions about code), and then heavy AI (agentic workflows)
I'll write a little tool that I can use to monitor my own metrics for some weeks with no AI tools (period), moderate AI (e.g. inline complete, asking questions about code), and then heavy AI (agentic workflows)
November 30, 2025 at 11:44 AM
but to "rethink"/clarify my own perspective with data gathered from my actual work.
I'll write a little tool that I can use to monitor my own metrics for some weeks with no AI tools (period), moderate AI (e.g. inline complete, asking questions about code), and then heavy AI (agentic workflows)
I'll write a little tool that I can use to monitor my own metrics for some weeks with no AI tools (period), moderate AI (e.g. inline complete, asking questions about code), and then heavy AI (agentic workflows)
finished @adamgrant.bsky.social's "Think Again", and it prompted me (pun intended) to take a more scientific approach to my thoughts on AI-assisted coding.
I have a number of theories, and I realized it wouldn't be all that hard for me to test them on myself, not to prove anything to anyone else..
I have a number of theories, and I realized it wouldn't be all that hard for me to test them on myself, not to prove anything to anyone else..
November 30, 2025 at 11:44 AM
finished @adamgrant.bsky.social's "Think Again", and it prompted me (pun intended) to take a more scientific approach to my thoughts on AI-assisted coding.
I have a number of theories, and I realized it wouldn't be all that hard for me to test them on myself, not to prove anything to anyone else..
I have a number of theories, and I realized it wouldn't be all that hard for me to test them on myself, not to prove anything to anyone else..
omg @lukeburns.com!!
I have so many things to catch up with you on, but for this thread what I was noticing is that most AI-assisted coding opinions I come across in the wild are steeped heavily in anecdotes and confirmation-bias (regardless of opinion), including my own.
I just...
I have so many things to catch up with you on, but for this thread what I was noticing is that most AI-assisted coding opinions I come across in the wild are steeped heavily in anecdotes and confirmation-bias (regardless of opinion), including my own.
I just...
November 30, 2025 at 11:44 AM
omg @lukeburns.com!!
I have so many things to catch up with you on, but for this thread what I was noticing is that most AI-assisted coding opinions I come across in the wild are steeped heavily in anecdotes and confirmation-bias (regardless of opinion), including my own.
I just...
I have so many things to catch up with you on, but for this thread what I was noticing is that most AI-assisted coding opinions I come across in the wild are steeped heavily in anecdotes and confirmation-bias (regardless of opinion), including my own.
I just...
Maybe i'll try this. I have some hypotheses on my own work but would have to actually run experiments tracking my own work to gauge.
November 29, 2025 at 5:45 PM
Maybe i'll try this. I have some hypotheses on my own work but would have to actually run experiments tracking my own work to gauge.
The best part is even if you don't end up finding a job tangential to that community, you're going to look look a hell of a lot stronger in any interview if you can start pointing to open source work you've done and speak to it.
November 28, 2025 at 6:16 PM
The best part is even if you don't end up finding a job tangential to that community, you're going to look look a hell of a lot stronger in any interview if you can start pointing to open source work you've done and speak to it.
You don't have to solve the big problems in some OSS project either. Just engaging in that community might lead to a job board for that tech.
Even if you don't land a job off of that, you get real world experience working with other engineers, which will give you experience and signal.
Even if you don't land a job off of that, you get real world experience working with other engineers, which will give you experience and signal.
November 28, 2025 at 6:13 PM
You don't have to solve the big problems in some OSS project either. Just engaging in that community might lead to a job board for that tech.
Even if you don't land a job off of that, you get real world experience working with other engineers, which will give you experience and signal.
Even if you don't land a job off of that, you get real world experience working with other engineers, which will give you experience and signal.
Virtually every job in tech I've gotten has been because I knew someone from working with them on a project/prior company.
I don't have data, but I'd bet it's easier today to find entry-level work like this than submitting dozens of cold apps on job boards.
I don't have data, but I'd bet it's easier today to find entry-level work like this than submitting dozens of cold apps on job boards.
November 28, 2025 at 6:13 PM
Virtually every job in tech I've gotten has been because I knew someone from working with them on a project/prior company.
I don't have data, but I'd bet it's easier today to find entry-level work like this than submitting dozens of cold apps on job boards.
I don't have data, but I'd bet it's easier today to find entry-level work like this than submitting dozens of cold apps on job boards.
This is what I came up with:
1. Pick a technology or problem area you like, ideally OSS. Go find their discord/community.
2. Build something in or with that. Ask questions, contribute to the project (genuinely).
3. Rinse and repeat.
1. Pick a technology or problem area you like, ideally OSS. Go find their discord/community.
2. Build something in or with that. Ask questions, contribute to the project (genuinely).
3. Rinse and repeat.
November 28, 2025 at 6:13 PM
This is what I came up with:
1. Pick a technology or problem area you like, ideally OSS. Go find their discord/community.
2. Build something in or with that. Ask questions, contribute to the project (genuinely).
3. Rinse and repeat.
1. Pick a technology or problem area you like, ideally OSS. Go find their discord/community.
2. Build something in or with that. Ask questions, contribute to the project (genuinely).
3. Rinse and repeat.