Introducing Donna: Your AI Second Brain App
I talk about my new app Donna, an AI second brain that lets you ask questions, add notes with auto-tags, and builds a memory profile of you. It uses Chinese open source models, allows data download, and I invite feedback.
Hear me out…
Two years back I joined X because I was tired….
Tired of jobs being posted on LinkedIn and within 1 hour of that, 100s of people had already applied. Everyone on LinkedIn was fake, trying to show how they were well-rounded and highly skilled. I barely made any REAL connections there.
Tired of tech influencers on Instagram saying, “COMMENT HI AND I WILL SEND YOU THIS IN YOUR DM,” just to increase their reach. I kept following them, commenting on their reels, getting the material in my DMs, and never really looking at it.
Then there was the cold email era where you send out 1000s of cold emails to recruiters, startup founders, and people in tech to either get a job or get some assistance for your startup. But let’s be honest, out of those 1000 emails, only 10–15 people (max) would reply.
But then there was X, where everyone is just being themselves. Even the most skilled people in your domain feel reachable here. Yes, there are a lot of shitposters, but there are also people who are really, really working hard and building something.
Hence, @shrav_10 and I are creating an X Chat for everyone interested in STARTUPS or TECH, where there will be only REAL tech talk.
We all can post important resources, job openings, and literally even collaborate and discuss startup ideas.
If you’re willing to actually GROW and NETWORK with the right kind of people here on X, then comment/quote/repost anything on this post and I’ll add you to the group chat.
Like I said, the right guidance and network can change your life forever!
Fable 5 is state-of-the-art on nearly all tested benchmarks, with exceptional performance in software engineering, knowledge work, scientific research, and vision.
The longer and more complex the task, the larger Fable 5’s lead over our other models.
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
A 2026 @KiaInd Seltos was presented to me for delivery with odometer tampered.
Dashboard: 20 km
OBD actual: 941 km
Dealer admitted driving it 750 km with meter disconnected. Said “very common practice.”
I refused. Dealer’s response: “we’ll give it to next customer.”
🧵👇
@jagograhakjago@nitin_gadkari@OfficeOfNG@MORTHIndia@Kia_Worldwide@KiaInd Dealer: NRL Kia, Greater Noida.
GM confirmed odometer disconnection on call.
Showroom management said: "If we knew you'd do a PDI, we would never have given you this car."
How many Kia buyers took delivery without a PDI and are driving tampered cars right now?
@jagograhakjago@nitin_gadkari@OfficeOfNG@MORTHIndia@Kia_Worldwide@KiaInd Update Day 13: @KiaInd marked my complaint as “resolved” without resolving anything. No refund. No acknowledgment of 941 km odometer tampering. Just closed the ticket. Had to get it reopened. This is the 4th time they’ve tried to close without acting.
CEO of Coinbase bragging that non-technical people are now using AI to ship production code. When a CEO says this, one of two things is true: either they don't know what production code is, or they're selling you a story.
This kind of hype is how businesses end up in real trouble. The people shipping the code and the CEOs cheering them on won't be the ones held responsible when it breaks. They collect their paycheck either way. Users will find out the hard way, once unreviewed code has touched their money, medical records, identities, private messages, kids' data, legal documents, critical infrastructure. By then it's too late.
Production code isn't a vibe.
The average person checks their phone 144 times a day, and each check costs the brain 23 minutes to refocus. Doing the math, most people are running at a fraction of their actual capacity and calling it normal.
The orchestrator doesn’t see the process. It only gets the final outputs.
Big shift:
Stop thinking in steps. Start thinking in independent tasks.
Because now you don’t parallelize code.
You parallelize agents.
Most people use Claude Code like autocomplete.
It’s actually an orchestrator.
It can spawn multiple subagents — each running independently with its own context and tools. 🧵👇
So instead of: Step 1 → Step 2 → Step 3
You get: Tasks running in parallel.
One agent reads code. Another writes tests. Another runs lint.
All at the same time.
This isn’t multithreading.
Each subagent is a full agent — it thinks, acts, and returns results independently.
We're bringing the advisor strategy to the Claude Platform.
Pair Opus as an advisor with Sonnet or Haiku as an executor, and get near Opus-level intelligence in your agents at a fraction of the cost.