I sometimes tweet about tech and career, other times about all random thoughts in adulting life.

Joined August 2013
mau kerja pakai spokenly malah sariawan
2
497
Folks in “i dont review PRs/code anymore” camp, Mind elaborating, how? IMO ai-generated code still depends on a) used model, b) how the dev worked on it At the end of the day, there’s bound to still be missed cases, no? Is it just an accepted tradeoff between speed/correctness?
7
27
1,609
I think I am starting to rediscover the joy in writing software in this AI age, though not by pumping out lots of code I've always enjoyed writing mini libraries, to learn how things work, now I am trying to understand how my brain works and make them somewhat deterministic
1
27
1,080
tired of chasing the hype of the always new and more capable models, currently more interested in how I can have a setup in which I can utilize non-frontier models to still do good works Feels like a more resilient skill to have compared to just throwing more AI subs at problems
8
177
some fun little addition to my personal site
6
316
Got on a slump, dropped all the way to 1350, now we're back here. Let's get to 1600 this time 💪
1
4
1,518
idk what's happening, climbed 300+ in a week
1
213
after ~3yrs of online chess, thousands of blunders, 1900 in bullet!
26
questions for engineering leaders a. Do you monitor token usages by your team members? b. Do you mentally (or even formally) use the metric along with other (e.g. task completed count) to make some classifications on your team? c. Any interesting thing you can share anonymously?
4
4
861
DIY @Steam machine + controller the controller only works with steam, which made me have to use steam remote play from the living room though apparently it's snappier than moonlight streaming, so all ends well 🙌
10
1,122
di umur 30 akhirnya bisa beli ioniq 5
7
1
38
3,966
TIL, given font-size: 14px line-height: 157.143% gives you 21.98px text height, but, line-height: 1.57143 gives you 22.00002px
231
using @googlechrome built-in local inference for generating mock data based on TS type defs still on origin trial tho, the model isn't exactly very capable but it seems sufficient for this use case developer.chrome.com/docs/ai…
3
282
GLM 5.1 seems to be promising enough for UI work without being as expensive as Opus This is one-shotted on z.ai chat interface with the following prompt "Build me a component library with a distinct memphis pop design direction."
2
222
2026 UX hacks
1
2
274
post-holiday blues is real
155
combat feels fluid world is beautiful almost perfect release timing, time to take a break
1
1
309
it's crazy to me that kotla.vercel.app (no update since 2022) is still getting traffic today slapped a service worker to it today, I need the 200k requests of my monthly free quota back
2
1
432
writing down my recent thoughts I wonder what will be different when I revisit it after a few months jackyef.com/posts/what-makes…
2
8
258
new peak, but I sense another lose streak back to 1600 incoming
1
248