Fadi Al-Majd

If life gives you lemons, make a battery out of them.

[CG @cgtwts]

Anthropic engineers:

“You’re not supposed to watch Claude Code work. You’re supposed to wake up and review what it shipped.”

One engineer built the entire workflow live in 22 minutes.

Most people stop when they leave their computer. Claude doesn’t.

Watch the video, then read the article.

[moontanax @xanaxmontanaonx]

YOUR NOTES ARE USELESS IF YOU CAN'T FIND THEM LATER

people save everything now:

> ideas
> books
> screenshots
> links
> quotes
> half-finished plans

and after 3 months it all turns into a folder nobody opens

this setup makes the whole thing feel alive:

> 3D graph shows how ideas connect
> quick capture saves thoughts before they disappear
> DeepSeek searches across the vault
> mental models apply to old notes
> Ask Vault answers from everything you saved

the interesting part is not the 3D animation

it's the direction everything is moving:

your files

your context

your own machine

AI that works closer to where your knowledge actually lives

[Fadi Al-Majd @GlobalFadi]

@alvinfoo A lot of people treat AI like magic. Under the hood it's tokens, embeddings, attention, and a lot of compute.

[Defileo🔮 @defileo]

🚨 One of the fastest growing GitHub repos in history just hit 66k stars.

Free GPT-4o, Claude, Kimi, Gemini, DeepSeek all under one roof, that's $2,400 back in your pocket every year.

One install, no paywall, no key, 66,000+ devs already know, you're still paying.

[Wes Roth @WesRoth]

Anthropic changed Claude’s dynamic workflow trigger word from “workflow” to “ultracode.”

The change is meant to reduce accidental activations when users mention the word “workflow” in a normal context.

[Fadi Al-Majd @GlobalFadi]

This is the right question. Not bigger robot models for the sake of it, but better data pipelines, evaluation, and learning after deployment.

[Yingdong Hu @yingdong_hu99]

#RSS2026 Call for Papers ✨
Workshop on Data-Centric Robotics: What Data Do Robots Really Need?

Share your latest work by Jun 20 AoE &
join us in Sydney with an amazing lineup of speakers from academia & industry!

https://t.co/iREG9BnwAi

@RoboticsSciSys

[Fadi Al-Majd @GlobalFadi]

@VraserX Source? Don't trust the hype. Verify the checkpoints. I'd still want to see real evals.

[Fadi Al-Majd @GlobalFadi]

@VraserX Reading about these new checkpoints makes me excited and cautious. The capability looks serious. Now let's see the consistency.

[Ejaaz @cryptopunk7213]

we’re almost at the point of no return.

anthropic has almost built an ai that autonomously self improves.. but there’s a catch

it sucks at taste. it’s terrible at picking research problems to improve. for that you’ll still need humans… well for now at least:

-> data shows newer claude models are 64% more likely to choose the correct next step in research.

-> it’s so concerning anthropic wants every lab to pause research before we (humans) lose control.

80% of anthropic code is generated by AI now, in 6 months my guess is this will be 90%+

despite all this i strongly doubt ai ever replaces the core elements of human taste. imo humans + models will always be the optimal combo

[Dipanshu Kushwaha @Dipanshu_AI]

🚨 Anthropic just showed a 27-minute workshop on how to actually do prompts for Claude.

Taught by the people who built it.

Free. No registration. No paywall.

I've seen $300 courses that don't cover what they teach in the first 8 minutes.

Watch it and bookmark it now.

[Fadi Al-Majd @GlobalFadi]

@KanikaBK This is the kind of AI breakthrough that makes me both excited and cautious. Recursive self-improvement is no small thing.

[Fadi Al-Majd @GlobalFadi]

@WesRoth I like the local first model. But don't ask users to trust it. Let them verify what stays on-device and what gets sent to the cloud.

[Wes Roth @WesRoth]

Perplexity is bringing Personal Computer to Windows, giving users a local AI agent that runs directly on their machine.

The rollout will begin with paying Max and Enterprise Max subscribers on the waitlist.

[Fadi Al-Majd @GlobalFadi]

@marclou Interesting how AI can one-shot feature logic, then struggle with the actual product feel. UI/UX is still very human.

[Hub71 @hub71ad]

The future of innovation starts here ✨

We're proud to welcome Cohort18 to Hub71, a new group of ambitious founders building solutions that will shape industries, solve global challenges, and drive the next wave of technological advancement.

Read more here: https://t.co/JnMvZmI1S1

#Hub71 #Cohort18 #InAbuDhabi

@persium_aq I @digistain I @matregenix I @LifeRubedo

[will depue @willdepue]

i never believed in the airgapped computer in a bunker fiction but i am slightly surprised to see how quickly claude has consumed anthropic and codex openai
again i find runaway RSI etc. whatever implausible but i would be sweating bullets if i did. not a ton of roadblocks there

[Fadi Al-Majd @GlobalFadi]

@k1rallik Reading this makes me both excited and cautious. More output is great, but verification matters even more now.

[chewa @chewaeth]

THIS 24-YEAR-OLD GUY FROM LISBON RUNS 6 BUSINESSES STRAIGHT FROM WHATSAPP. NOT A SINGLE LINE OF CODE. 5 SCREENS. $63,000 A MONTH

He never opens an editor. Never touches a terminal. Never hired anyone. He just types into a chat and the agents do the rest

Right now one screen has an agent building his new product https://t.co/Gm86ktKDvo from scratch. He drops a voice note, the agent writes the code, he hits "ok" and scrolls on

5 monitors, 5 businesses, each one running on its own. Every one has its own agent. All the tasks land in a single chat

While everyone else is copy-pasting boilerplate and arguing about frameworks, he's approving everything from his phone lying on the couch

No office, no team, no coding skills. He says what he wants and it shows up

Six months ago he was a solo freelancer. Now he's got 6 products and zero employees

He calls it the top skill of 2026. And looking at those 5 screens, it's kind of hard to argue

[Fadi Al-Majd @GlobalFadi]

Software just changed. To me the bottleneck is no longer writing code. It's judgment, verification, and deciding what is actually worth building.

[Riley West @rileywestreel]

🚨 do you understand what just happened to software Anthropic..

Anthropic's own engineers now ship 8x more code per quarter than they did in 2021–2025. The reason: Claude writes over 80% of the code going into Anthropic's codebase. A year ago that number was low single digits.

The metric that matters is task length — how long a job the model does solo without a human. It doubles every four months. March 2024 it handled 4-minute tasks. A year later, 90-minute ones. Now, 12-hour ones. "In a year Claude went from very useful to superhuman."

→ On a standard training-code speedup test, May 2025 the model hit 3x. By April 2026, 52x.
→ A human expert needs 4–8 hours just to reach 4x.
→ On an autonomous AI-safety study, agents closed 97% of a research gap in 800 hours for $18,000 in compute.
→ Two humans got 23% in a week.
→ April 2026: Claude shipped 800+ fixes a human engineer estimated at 4 years of work.

The 99% perspiration Edison talked about just got automated. Taste is the last thing left, and taste is just another skill the model hasn't learned yet.

[Fadi Al-Majd @GlobalFadi]

@zeuuss_01 The beauty of a local agent lies in its predictable cost and control. For daily workflows, that is a stark contrast to the usual API treadmill.