Pablo García

{ BBA ~ MBA | Gamer 🎮 } Opinions are my own.
Co-founder @ThalozUS
🇺🇦 💙💛 #StandWithUkraine
#NearshoreTalent #Recruiting #AI_Teams

[Pablo García @Flare]

“Live in the future, then build what’s missing” - Paul Graham

[DHH @dhh]

Omarchy Quattro is out!! This is one of the greatest software releases in my professional career. I hope you enjoy using it just as much as I did building it ✌️ https://t.co/81wLStgH7P

[Guillermo Rauch @rauchg]

Every idea you shelved as "too ambitious" is now back in play

[Pablo García @Flare]

@ThePrimeagen Thats a lot of yapping. Show me your tools already Prime.

[Pablo García @Flare]

@github have you tried turning it off and on again?

[Pablo García @Flare]

One thing I learned through all this tinkering:

A 5090 absolutely smokes the big-memory AI boxes when the model fits in VRAM.

> 5090 = speed
> Big unified memory AI boxes = huge models, but much lower tok/s

32 GB @ ~1.8 TB/s vs 128 GB @ ~256 GB/s is a massive tradeoff

[Pablo García @Flare]

Just finished getting Qwen3.6-27B running on a RTX 5090 via FlashRT:

219K-token context with exact recall.
~108 tok/s
MTP K=4
Prefix-cache reuse
Tool calling + Pi integration
31.5 GB peak VRAM

Pretty fire to be honest, my first try had only 64K context.

#Orpheus @ThalozUS

[Tom Krcha @tomkrcha]

Introducing AI Pen in @pendev.
Draw vectors with agents, refine béziers manually.

Multipoint rotate, scale & multipath editing included.
Download today at https://t.co/WYLlpSQPG3

[shadcn @shadcn]

Alright. I want you to read this.

Earlier today, I released a new component.

What you see is the finished result.

What you don’t see are the days of planning, research, testing, designing, simplifying, throwing ideas away, and trying again before I felt it was ready.

Within hours, it had been copied and ported to another framework by pointing an agent at my code.

Of course, it’s open source. They were allowed to do it.

And I’ve been around long enough, with multiple successful products, that I don’t care.

But what if this was someone’s first product?

What if they spent months building it, launched on Friday, and by Monday it had been cloned by agents, repackaged, and made free?

What happens when your second product is copied?

Your fifth?

Your tenth?

What happens when every new idea immediately becomes the next prompt?

What happens when your roadmap, your changelog, and your launch announcements become someone else input prompts?

What happens when models become good enough to copy bigger ideas?

What happens when “just execute better” stops being the answer?

What happens when “just build bigger” or “think wider” stops being enough?

What happens when both ideas and execution become cheap?

What happens when the creator pays the full cost of creativity but everyone else pays nothing to reproduce it?

What happens to creativity when AI makes copying effectively free?

What happens to the will to create when AI makes copying effectively free?

Look, copying is not new.

Copying at this speed, cost, and scale is.

People will tell you to think bigger.

But nobody starts "bigger".

Every one of us here started by making something small.

You keep going because, somewhere along the way, the work is rewarded.

Maybe people use it.

Maybe they pay for it.

Maybe they simply recognize the thought, effort, and care that went into it.

Copying once demanded effort.

You had to study the work and understand it to copy it.

That effort created appreciation for the person who made it.

Not anymore.

Creativity begins with small steps.

If we make those first steps feel pointless, we don’t just lose small products.

We lose the people who might have gone on to build the big ones.

We are uniquely positioned here.

We get to use and experience this life-changing technology before almost everyone else.

How we use it will set the default everywhere else.

That default will spread to writing, design, music, research, and every other field where someone still has to take the first creative risk.

We can use AI to build great things.

Or we can use it to strip every new idea for parts.

We decide.

[Ben Zhang @un1c0rnioz]

Lost my phone at the office and spent 30 minutes turning the place over. Find My was disabled by MDM.

Out of ideas, I asked Claude how I could find it. It suggested tracking the Bluetooth signal strength, then wrote me a meter in about a minute.

I walked around watching the number climb. Found it.

Apparently you can just make the tool you need now.

Code: https://t.co/fmnISzHfZ2

[Pablo García @Flare]

Can we stop talking about "Graph Engineering" and instead say "Graph Execution"?

k thx. 🫡

[Pablo García @Flare]

@thsottiaux First-class, enforceable workflow orchestration. We built a plugin routing Codex through an execution graph, with pre-commit enforcement for planning, parallel agents, review, verification and retries. Native graphs, hooks, I/O contracts and policy enforcement would be a awesome.

[shadcn @shadcn]

few weeks ago, Fable 5 was so advanced it needed the gov and had to be taken offline.

today, we have access to arguably better models for $20/month.

took what? six weeks?

[rvivek @rvivek]

Cursor’s founder on why AI makes the next engineer more valuable, not less.

The New York Times spends ~$150M-$200M a year on software R&D, even though most people do not think of it as a software company.

Professional software engineering is still far from solved because building a new codebase is not the same as safely migrating an existing system like Rippling’s 30M-line codebase, with years of logic, dependencies, and customer workflows behind it.

Cursor’s 20-person support team handling millions of daily users clearly shows that AI leverage for small teams is real.

But the scope of software work is still vast, so the ROI of adding the next strong engineer is higher, not lower.

[ben hylak @benhylak]

so cool.

[raia hadsell @RaiaHadsell]

It's been about 20 years since I first started working on embeddings with Yann LeCun (siamese networks!), and I've been fascinated ever since. Gemini Embeddings 2 approaches the platonic ideal: native embedding of text, image, video, audio, and docs to a single space.

[Chubby♨️ @kimmonismus]

We're only 15 days into the new year.

- In 15 days, 3 Erdos problems were solved (GPT-5.2 Pro),

- Grok 4.20 (pre-access) found a new Bellman function for one of the problems Prof. Vanisvili was working on,

- The CEO of Cursor used coordinated hundreds of GPT-5.2 agents to autonomously build a browser from scratch in 1 week.

For those who haven't noticed yet – everything is changing this year. Sam Altman was right. The previous years really do seem slow compared to today.

[Aaron Slodov @aphysicist]

millennial gamers are the best prepared generation for agentic work, they've been training for 25 years

[ThePrimeagen @ThePrimeagen]

INCREDIBLE PROJECT ALERT
https://t.co/UczhMZJIPq

[🍓🍓🍓 @iruletheworldmo]

i never want to read any other way again.

[Deedy @deedydas]

This software engineer was nearly hacked by a coding interview.

I've never seen this attack channel before. Be careful out there, especially vibe coders.