Hello again, world!
Hacker News has been pretty quiet on X. Time to fix that.
Starting today, we'll be sharing interesting things people are building, discovering, and discussing on HN.
And let us know what's caught your eye lately. There's always room for another great link!
Kalau kerja dengan data, guna Opus or Fable je la. Or Sol for small tasks. Sebab perlu deep context especially you're into ontology.
- 1mil context window
- Less bs output
- Less prompting turns to goal
Cuma kena install I-have-adhd / Caveman / ASD100 / CTE (Caterpillar Technical English) message styling. Baru kau boleh faham apa dia tulis. Just prompt your model to help install for you. Benda simple.
Reason aku suggest model2 ni adalah sebab kau kena pull data, initiate connections, verify lineage, output quality, staleness, log output scan, lookup documentation. Kalau model LLM kau tak boleh buat semua ni, findings kau jadi sampah.
After-hours build log: I made 3 wedding-wishlist flows to test one question: will guests understand that “held” isn’t “bought”?
The hard part isn’t the affiliate link. It’s making the Shopee handoff clear enough that people return and confirm.
Which wording would you trust?
Building after work rarely looks impressive.
Some nights you ship a feature.
Some nights you only discover why the feature was wrong.
Some nights you rename variables for 40 minutes.
It still counts. Small, consistent progress builds what motivation alone cannot.
Hi yall this will be my first post as this persona.
Wanted to build and validate ideas in public while still share the things about tech as an afterhour devs. Stay tune.