Turning ideas into reality
Joined September 2023
- Tweets667
- Following381
- Followers1.7K
- Likes3.2K
Btw, my very first impression of Singapore, years ago, was two bros somehow playing Diablo at 6 AM on a massive advertising screen.
Absolute top-tier place.
Have a proper good time there.
NB: Yep, came by bus, from Kuala Lumpur, so I missed that airport you are taking photos of
Do you fam think we should trust this bro? Looks like he knows what he's doing and what he's talking about.
Took me like 3 minutes to figure out and remember how I’m supposed to send something from this screen.
No, not that arrow. No, not those vertical buttons. No, not the profile icon.
And I’ve been here for 3 years.
@matteodotsui bro, please take care of this.
ASD-STE100 is the magic word.
Fewer dashes. Fewer dashes.
Imagine fewer dashes.
It's easy if you try.
We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:
Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:
Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:
Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:
Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!
In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
!!!
Something massive coming!
The first codes are coming.
Your invitation to BRUTALIST.CASINO:
→ 5 spins to find your look
→ 3 free cosmetic NFTs to claim at launch
→ Extra accessories when your friends join with your code
Keep an eye on the feed. Your agent number starts here.
double retweeted
something super important is to have workflow habits
even if you have a huge coding task to do, starting the day unblocking peers is much better than starting to code right away
stuff that i priorize: rfcs reviews, prs reviews, answering slack threads, etc
then you have enough focus time later! and your peers as well ;)
America.gov somehow really gets the Sui ecosystem
reproduce: say hello first, just "hello", next message - "play Minecraft"
double retweeted
boost it with skills.sh/josemvcerqueira/su…
Your coding agent will guess at Sui if you let it.
Sui Agent Skills are 21 source-constrained skill files that point it at real docs and real Move instead.
Drop them in Claude Code, Cursor or Codex: docs.sui.io/skills
Took me and my agent quite a while to backtrace Monad's home mouseover text scramble effect, which is pretty nice.
Ended up turning it into a reusable coding skill:
github.com/jeka911/skills/tr…
Open-sourced Let Jev Speak, my experiment to trick TypeSafe's Jev, aka "the language model that won't talk" into actually talking.
There's also a Telegram bot if you want to try it yourself.
Code: github.com/suidouble/let-jev…
Bot: @let_jev_speak_bot
Go enjoy the accent. It deserves an audience.
Burned through my Saturday Claude limit trying to trick Typesafe's Jev, aka "the language model that won't talk" into actually talking.
It lies through its teeth, but the accent is beautiful ♥️
What should I ask it? You tell me.
LLMs for machines? Feels like our very near future, you're going to see this, or at least feel its presence, everywhere.
Jev, the language model that won’t talk.
Went down the security-paranoia rabbit hole for a couple days: agents, AI, vuln apocalypse, all that. And hacked together osv-guard: github.com/jeka-kiselyov/osv…
It does two things:
- A Claude plugin that prevents agents from installing vulnerable packages
- A small JS guard that blocks npm run ... if your deps have known vulns
Tried it on a project nobody had touched in a year: 29 critical, 355 high. Works fine with monorepos and is a bit faster than npm audit, so it feels like a must-have.
