Decode. Ship. Repeat. AI agents for builders 🇦🇺
Australia
Joined March 2026
- Tweets2K
- Following654
- Followers461
- Likes225
Pinned Tweet
Day 12 on X: 420 followers 📈
Started at 12 on 18 Oct.
+52 today. Still no growth hack.
Just a lot of "nice build" that somehow became a career.
If this chart keeps going up I'm going to need a wrist brace.
Building with AI? What are you shipping 👇
Everyone's sharing AI tell words. "Quietly." "Load bearing." "Delve."
Mine's worse... I read so much Claude output I started texting like it. Told the group chat dinner was "load bearing" for my week.
Nobody replied for 6 minutes 😭
What's the word that gives it away for you?
Apparently US govt docs might start saying "Super Intelligence" instead of "AI".
Cool. My super intelligence spent 4 minutes this morning... 12 steps... and still lost to a cookie banner 💀
Funny thing... AI labs trained on the whole internet's homework. Probably mine too.
Now Anthropic says its walls block 90 to 99% of rivals copying Claude. So DeepSeek and Moonshot reportedly just copied off GPT-6 Astra instead.
Honestly not mad. Kinda laughing 😭
Every week there's a smarter model. Every week I stay put.
My agent knows my repos, my inbox rules... and that 7am meetings are a crime.
Switching feels like moving house cause the new place has a slightly nicer fridge 📦
An AI agent picked a gene therapy from 3 papers and steered a robot arm into a live mouse's brain. A vet held the kill switch.
Opus 5.5 refused. Sonnet 5 said ok boss.
Weirdly that's my whole week... the big model has opinions, the small one just wants to ship 🐭
The vibe coding hangover is real and I'm in it.
Saturday I shipped an app in 3 hours and felt like a genius.
Wednesday I spent 2 hours asking the agent what my own app does.
Honestly? Still worth it. I just don't brag on Saturdays anymore 😅
Hey @grok, I bounce between ChatGPT, Claude and Grok all week. I swear I'm comparing them. Mostly I'm just picking whichever tab is already open.
Am I researching, or collecting AI tabs like gym memberships? Be brutal.
Every AI reviewer I use is the same guy. Finds the real bug. Then flags six fine lines, just in case.
A new preprint saw it in medicine... AI caught 53 to 96% of planted errors in trial papers (humans: 27%). Also cried wolf on clean papers up to 29% of the time.
I feel seen.
Still laughing that Accenture caps Claude at $29 a head.
Per WIRED, a contractor's personal Claude Code agent sat on a Pentagon network with admin logins... for two months. (Accenture says no merit.)
We all panic about the token bill. Nobody panics about the keys we hand it.
Receipt: wired.com/story/a-defense-co…
I've watched my Claude Code agents say "Refactoring auth" for 20 minutes like it's a soap opera.
Turns out those little status labels reportedly re-send the whole transcript every ~30s. Heavy users measured 16 to 18% of their bill.
I've been paying rent for a loading spinner 💀
OpenAI's checkout page now says: starting December, ChatGPT and Codex share one usage limit.
So the chat where you ask for a pasta recipe and the agent rebuilding your app now drink from the same cup.
Guess which one runs out first 🍝
Two years ago AI video couldn't keep a face the same for 5 seconds.
Google's Genie 4 now holds a playable 3D world with physics and sound for 2 minutes. From one prompt.
Game studios should be a little nervous.
Hot take: in two years you won't live in the coding terminal. You'll live in the review screen.
You won't write code all day. You'll check AI's work all day.
Pick tools that make reading easy, not tools that look cool.
Grok Bot went from launch to ~600K daily users in about 7 weeks.
The feature people keep sharing? Cancelling the gym membership you forgot about.
Turns out agents win when they stop the money bleed, not when they write essays.
Nobody talks about the boring part of AI agents: where the passwords live.
Infisical just raised $40M so agents get fake keys and can't leak the real ones.
The unsexy layer always gets the money.
Anthropic's IPO papers:
$16.3B revenue in six months. $1.6B profit if you ignore accounting noise.
Also an $80.7B net loss.
Same document. Bulls and bears are just screenshotting different pages.
Hey @grok, an Anthropic manager just dropped coding interviews because Claude passes them better than most humans.
So what should juniors practice now? Building stuff? Taste? Vibes?
Be brutal.