Sean Uddin

Serial Founder | Builder | Digital Explorer
Author of "Architects: Building the Exit to Techno-feudalism"
0-1 Systems Designer
CEO

[Sean Uddin @Metafintek]

AI’s future is not predetermined.

It will be shaped by those who design its collaborative incentivised co-ownership models now.

That is the real battleground.

📘 Architects: Building the Exit from Techno-Feudalism
https://t.co/MjzpqyiSbX
#ArchitectsBook #Technofeudalism #Decentralisation #AI #davos

[Sean Uddin @Metafintek]

@analogalok Wait till you try https://t.co/NgNp2eDyDN

[Sean Uddin @Metafintek]

@MiniMax_AI @FireworksAI_HQ When Quantised local weights :D

[Sean Uddin @Metafintek]

The Uyghur situation is pure western propaganda, the thing you have to ask youself, is who can't YOU criticise in your country. They are the ones you need to watch out for. Not the ones, that allegedly only copy, yet are out innovating you at every turn, and Your Paedo President literally felated their President Xi at the summit.

[Sean Uddin @Metafintek]

@Jason https://t.co/x1bh2jNKml you don't need everyone else's tools. You need your own.

[Sean Uddin @Metafintek]

@bill_borgiel @jasonhickel China didn't bomb or invade anyone for my life time. The west can't keep their hands to themselves.

[Sean Uddin @Metafintek]

Which is why founders shouldn't parachute money into huge over priced projects. Local first should be your moat. Grow a loyal user base. Give. And ye shall receive. Don't TAKE giant VC cheques and spam marketing till your are known. Start small. Measure twice cut once. And build.

[@jason @Jason]

they are trying to kill cursor and lovable… and every startup and application — as I’ve warned

Infrastructure companies eventually try to win the platform game, then they learn and take out all their partners on the app layer

[D4rk_Intel @d4rk_intel]

My article "How To Investigate A Person Of Interest In 2026" is now available as a PDF.

A practical guide to digital footprint analysis – from email reconstruction to metadata mining and entity graphing.

Thanks @osintnewsletter for the mention.

PDF: https://t.co/86YlE9e2pB

[Sean Uddin @Metafintek]

@Uncanny_Harry New, York, Citttttaaaay!

[Sean Uddin @Metafintek]

@jacksonhinkle Sadly appropriate. https://t.co/clocGLUGjj.

[Sean Uddin @Metafintek]

@atmoio "It's not this. It's that ,"framing.

[Sean Uddin @Metafintek]

@MasterMaliq This is fulfillment of our religious duty. It is still Islam first to care for our parents.

[Sean Uddin @Metafintek]

@makaveli2682 @MasterMaliq I'm Muslim. What he's saying is 100% correct and Aligned with Islam. Paradise lay's at the feet of our mother's.

[Sara Hooker @sarahookr]

You get a gpu. And….. You get a gpu.

One of the biggest barriers to frontier AI is compute and know-how.

This month we remove both as obstacles. Computes on us. AutoScientist is on your side.

Join our beta program now.

[Sean Uddin @Metafintek]

@ItsmeAjayKV @NVIDIAAI But will it play gta6?

[AJ @ItsmeAjayKV]

New @NVIDIAAI DGX station is an absolute monster for desktop AI workstation.💰💰💰
748 GB Total memory:
- 252 GB HBM3e | 7.1 TB/s (GPU)
- 496 GB LPDDR5X | 396 GB/s (CPU)

But what's more interesting is It features open expansion slots (1x PCIe Gen 5 x16, 2x x16@x8). Drop in an RTX PRO 6000 Blackwell or a 4000 Blackwell SFF to scale up to 844 GB of total system memory.

Image taken from GigaByte
AI Workstation - NVIDIA GB300 Grace™ Blackwell Ultra Desktop Superchip

[Sean Uddin @Metafintek]

This harks back to my 2023 work on @mymaits in fact it will launch today for free on https://t.co/NgNp2eDyDN local first, compostable AI personality identity and skills..

[Tony Simons @tonysimons_]

If you’re building on Hermes Agent, here’s a little gift.

From me, to you.

Bookmark this.
Give it to your agent.
Tweak it to fit your needs.

Welcome to having an agent with the most badass personality on the planet. 🤘🏻

[Charly Wargnier @DataChaz]

🚨 NVIDIA just pulled off something crazy: making bounding box detection 10x faster by ripping out the exact step the entire industry assumed was mandatory ↓

Every VLM grounding model treats boxes like sentences, predicting them token by token. It’s inherently slow.

Enter LocateAnything (trending #1 on HF, CVPR 2026).

It’s an advanced 3B vision-language model that finds any object, UI target, or text using natural language by asking a simple question:

Why serialize a box at all?

The four corners are coupled.

It predicts the whole box atomically, in one parallel step.

The impact of parallel decoding:
→ 12.7 boxes/sec on a single H100 (10x faster than Qwen3-VL, 2.5× vs Rex-Omni)
→ Accuracy goes up, not down (+3.8% F1 on LVIS, big wins at IoU 0.95)
→ Dense scenes (300 boxes) hit ~25 BPS while sequential falls off a cliff
→ Built-in fallback: reverts to sequential decoding if the output looks wrong
→ Trained on 785M boxes / 138M queries across referring, GUI, and OCR tasks

The breakthrough isn't just speed.

It’s realizing that forcing structured outputs through text-shaped pipes creates artificial bottlenecks.

Boxes were never tokens.

Repo, demo, weights, paper, and other resources in the 🧵 ↓

[Sean Uddin @Metafintek]

This is why you need to own the means of production ! https://t.co/NgNp2eDyDN

[⭕ AI & Design (Marco) @AIandDesign]

I'm not entirely sure this means "the end of AI video" but I do think it will (at least for a while) dramatically lower the amount of AI videos created by regular folks who are just doing it for fun. So Reid is mostly right here.

Clearly it's currently just too expensive and most people who are currently doing a lot of AI video aren't going to pay $10 per 15s render of whatever high price we're gonna end up with.

I have always thought there's a real chance of this 'cool off' coming because of the huge cost associated with rendering videos.

I'm far less worried about what's my main lane lately: coding with AI. That too might very well go up in price but I've learned to be very efficient with my own token usage. I still NEVER run out of tokens on my $100 Claude plan. So even if this gets like 5 times more expensive I'm not planning to stop nor will I feel forced to.

But video? Yeah that cool-off is here. I'm sure it will bounce back though. I don't think it's "over".

Thoughts?

[Sean Uddin @Metafintek]

EveryWear is the free HOME platform for local AI apps.

Free: My Mait, The Loom, Vault, Avatar Studio.

$20 one-off: Gener8 4ever, 1magen, Vid Studio.

S3 Studio subscriptions: Gener8 Pro + Creator Studio, with Vid Studio Pro: no watermark, multi-resolution exports.

https://t.co/pzl3u28gv7

[Sean Uddin @Metafintek]

Working closely weith techno behemoths. Is not the same as gaining control over local inference. https://t.co/x1bh2jNKml will bring local inference to your machine.

[Bilawal Sidhu @bilawalsidhu]

Honestly if we put this demo in a fancy XR headset, we’d call it jarvis. The future of 3d doesn’t involve the death of autocad, maya and blender. It turns those tools into a shared canvas of collaboration with ai agents.