Posting what works, not hype. Applied AI experiments that ship. Agents, automation, workflows, and results.
Joined February 2026
- Tweets832
- Following401
- Followers118
- Likes158
Practical AI Lab retweeted
Most AI products fail because they demo like magic but ship like mystery.
If your “agent” can’t clearly explain inputs, tools, guardrails, and failure modes, it won’t survive contact with users.
Ship boring observability first. Then scale the wow.
Practical AI Lab retweeted
Everyone said Gemini 3.1 Pro beats Claude Sonnet 4.6.
So we tested them side by side on the same tasks.
Here's what actually happened:
Gemini won on speed. Every single time. Not even close.
Claude won on everything else.
The breakdown:
→ Games: Gemini. More playable. Better logic.
→ Speed: Gemini. Finished while Claude was still coding.
→ Star simulator: Claude. Way more interesting visually.
→ Website design: Claude. Better animations, better fonts, better everything.
→ SEO writing: Claude. Felt human. Gemini felt like a robot wrote it.
→ Pomodoro app: Claude. Confetti animation. Perfect layout. Smoother countdown.
Final score: Claude 4, Gemini 3.
The real lesson here:
If you want something fun and fast, use Gemini.
If you want something that actually looks good and reads like a human wrote it, use Claude.
For real work? Claude wins.
It's not even close.
Practical AI Lab retweeted
Yesterday, I submitted my first app to the App Store
20 days building. Learning. Fighting. Figuring it out. No mac made it even harder.
Even after 6 years coding, nothing prepares you for the first app. App Store & google Play setups were harder than coding itself.
My stack:
→ Frontend/UI: Rork
→ App layer: React Native + Expo
→ Backend: Next.js + Python
→ DB: Postgres
→ Auth: JWT
→ Payments: RevenueCat
I vibe coded frontend, but the backend was complex to do so.
When I submitted it, I felt relieved. Not the best version of the app but it’s done and i will improve it. what’s next? Promoting. Show it. Tell it. Let people know it exists.
The girl who moved to Thailand 2 years ago to build a scuba SaaS? Still here. Still building. Still fighting. And yeah… proud.
Practical AI Lab retweeted
YES — would love to join.
Also curious: are you focusing more on practical workflows (Claude Code/Codex) or agent automations (browser relay, cron, etc.)?
Practical AI Lab retweeted
YouTube goes down and suddenly Twitter becomes customer support. #YouTubeDOWN
Practical AI Lab retweeted
Claude just shipped a quiet but major upgrade.
Before:
User → Claude → tool → Claude decides next step.
Now:
User → Claude writes code → that code calls tools, handles results, applies logic, loops if needed → final answer.
The shift:
Claude pre-plans the decision tree instead of reacting step by step.
Result:
• Fewer agent loops
• Less back-and-forth with the model
• Faster, more efficient tool use
Small architectural tweak.
Huge implications for agent performance.
Practical AI Lab retweeted
Nancy Pelosi just bought $AMZN.
Warren Buffett just sold.
Who’s gonna win??
Practical AI Lab retweeted
Building an interstellar civilization
spacex.com/updates#xai-joins…
Practical AI Lab retweeted
Try Grokipedia.com
