đȘđž Spaniard. Entrepreneur. Dad. đ„ Miami Heat fan @handcashapp (exit 2024)
Andorra â Dubai
Joined July 2009
- Tweets74.9K
- Following2.5K
- Followers9.4K
- Likes121K
Pinned Tweet
Two years ago I sold my company.
I thought the next chapter would be simple. It wasnât.
Iâve been in Andorra since 2021. Itâs beautiful, my daughter was born here, and it will always be special.
But with ~80k people (mostly freelancers and retirees), no real network, and the nearest airport 3 hours away, itâs not the place for someone in their 30s who still wants to build.
I aimed for the US but couldnât get the visa (long story involving a law firm in Florida).
So, after much consideration, I'm moving to Dubai.
Iâve been an entrepreneur most of my life.
Right now I want to join something thatâs already big, in motion, and fun. I donât want another remote-only chapter.
I want an office, a team I see daily, big challenges, and high standards.
It took me a while to admit that staying here meant I couldnât become who I want to be.
I canât stand that any longer.
At HandCash (CEO & co-founder) I ran product, design, and marketing. I raised money, hired, and led teams.
We tackled some of the biggest challenges in crypto/fintech and Iâm proud of what we shipped.
The team came up with some of the most relevant innovations in our sector.
Iâve been deep in AI since before it was cool. In 2016 I led Accentureâs first AI division in Madrid for a year (until we raised âŹ1M for HandCash!)
People whoâve worked with me usually say Iâm honest, and my former employees always said I was a great leader. (Don't take my word, ask them!)
If you know a founder or executive in Dubai building something cool in AI, robotics, or fintech, please share this with them.
Iâd love to have a coffee in the coming weeks. I land in Dubai on Nov 10th.
DMs are open or email me at [email protected]
If youâre reading this, help me by sharing it.
Thanks.
PS: This is a photo of me and Rafa, my co-founder in multiple startups! 2018... this is the biggest office we had lmao! I think you get it now.
Alex Agut retweeted
Made a fun little Mr. Sandman quartet with @jamesonhodgeâs CHIMERA.
Every voice is generated as real-time singing from a synth simulating vibrating vocal folds and sound waves travelling through changing mouth shapes.
Alex Agut retweeted
This is what building from first principles looks like.
Alex Agut retweeted
Andrej Karpathy:
âA 1B model can be much more effective than most current LLMs.
90% of current models waste a ton of capacity and money on remembering stuff that doesnât matter.â
here is his 5 core thesis about Small LLMâs:
1. the intelligence was never the knowledge. Strip the memorized internet, keep the reasoning. Whatâs left is the cognitive core, and itâs small.
2. the core might fit in ~1B params. Maybe less. âPlenty of room at the bottom.â Weâve been paying for a library when we needed a brain.
3. size stopped being the bottleneck. What makes an agent work is the tools it can call and how tight the feedback loop is, not parameter count.
4. intelligence lives in the system, not the weights. A raw LLM is a CPU with no OS. Small core + good harness > giant model alone.
5. you can own it. A model on your hardware is one no one can rate-limit, price-hike, or deprecate on you.
in 12 months the gap between teams shipping on-device and teams paying per-token frontier bills wonât be capability.
TwIL an 8.8B model at 90%+ of frontier isnât a benchmark flex.
Itâs the thesis Karpathyâs been repeating all year, made real.
study more on this below â
Modest intelligence is arriving faster than most people realize.
Today, we're open-sourcing TwIL-8B.
Just 8.8 billion parameters, yet approaching frontier-model scores on several difficult benchmarks:
âą 93% of frontier performance on coding
âą 92% on instruction following
âą 90% on contest mathematics
All from a model small enough to run on your own hardware.
But the benchmarks aren't the most important part of this announcement.
For years, the AI industry has focused on building the most capable model, regardless of the cost required to run it.
I believe that equation is changing.
The next era of AI will be defined by the cost of completing work. Not simply the intelligence of the model, but how efficiently that intelligence can accomplish the task.
If a model with 8 billion parameters can reliably complete the same work as one with hundreds of billions, why pay for the larger model?
As smaller models become more capable, intelligence becomes something we can distribute, specialize, and run across the hardware we already own.
That's the future we're building at @thewebAI
TwIL-8B is available today. Open weights.
And this is just the first of several announcements we're making today.
HF Link: huggingface.co/webAI-OfficiaâŠ
Alex Agut retweeted
This is the future.
I love these sort of new high-tech industrial companies popping out everywhere! This is the way!
At twentyfour26, we are building fully autonomous factories in America, starting with on-demand CNC machining, manufacturing components for hardware companies in and around the Bay Area, with turnaround times as fast as ~48 hours.
Automation and robotics have made manufacturing more efficient, but much of that progress has come from optimizing processes designed decades ago.
We believe the next leap in manufacturing wonât come from simply automating existing processes. It requires rethinking how factories operate from the ground up, with AI at the core of the entire production process.
We need to reinvent manufacturing.
Get out of the European Union as fast as you can.
x402 is one of the best things right now and nobody uses it. Probably nobody really gets it.
In short, you make an API request with the payment embedded on it. That way you can use paid APIs without having to sign up for every new service.
This is the true on-demand economy.
Alex Agut retweeted
if you're a founder looking to pivot or start something new, don't build another sandbox agent provider. start a construction firm and make it ai native. you'll build a better business and create way more value for the world
everyone is crowding into the same few ideas, while the rest of the economy is basically untouched
most businesses in the world are old. they were built before software, and most never really adopted it. ai changes that, because it makes both great operations and technical execution much easier than before
so the opportunity is no longer selling tools to these industries. it's rebuilding them from scratch, ai native from day one
new construction firms. new schools. new hospitals. new clinics, logistics companies, property managers. every old industry is waiting for someone to rebuild it
in ai infra you compete with 50 funded teams shipping the same thing every week. in a boring industry you compete with companies that haven't changed in 25 years
pick one, obsess over it, and learn the business better than the people in it. the technical part is the easy part now
the time has come to improve everything
Alex Agut retweeted
I really don't understand this.
I see many people now apply to @ycombnaitor with the app that they vibe-coded (some even with revenue).
It makes no sense to me - even if you get in, money won't save you- and this is why I never applied to YC with Postiz.
I would never invest in a vibe-coded app, and I got offers from startups making above $1m ARR (built pre-AI era, of course)
Don't get me wrong, it can make a lot of money (Postiz: $4.3m ARR), but the chances are slim.
Just build it yourself and get it to $10-$20k MRR.
When to apply to YC? When you have
- Deeptech startup
- Medical
- Security
- Infrastructure barrier (think Deel, Stripe, etc.)
This isn't to bum you. It's to save you.
If you raise a lot of money, you would have to make a lot more money- bottom-up is not an option anymore. You need Enterprise deals, and a lot.
You'll also be flooded with competitors because everyone can build what you build. So you will need to be reaaaallly good at marketing.
Your salary will be capped by investors, and in the end, you'll most likely close your startup. Your risk of success becomes much, much lower.
Don't get me wrong, I am not the - you have to go bootstrap. I am not. My next company will most likely be VC-based.
After Postiz, when I am rich already, I don't care to sacrifice years of my life for VC and build something huge (not vibe-coded, of course), probably deep tech.
Alex Agut retweeted
đšATTENTION - HandCash MNEE holders.
Convert your MNEE before Dec 31 (no extension).
Email [email protected] with the amount held in your wallet, and a corresponding receiving address for USDC/BTC/ETH/BSV.
Theyâll send you a unique deposit address to send your MNEE.
Alex Agut retweeted
What happened (I know nothing)
- bot launched
- 2 weeks later usage flatlined
- Elon asked whatsup
- PM says itâs your model, product is fine
- Elon says ok, use whatever model
- Team switches to Opus 5.5
- Growth skyrockets
- Elon has never seen anything like it
- see Anthropic bill
- đ€Ż forces quota cut
-> you are here
mfw when the model is the product
Alex Agut retweeted
GENIUS đ€Łđ€Łđ€Ł
He created a script to counter the new data tax the French government wants to adopt:
- He downloads a PDF file available on the Socialist Party's website on an infinite loop, which racks up âŹ30 in taxes in 4 hours.
- In theory, if 1,000 people did this, they could drain their âŹ16 million budget in 5 months đ€Łđ€Łđ€Ł
The law doesn't exempt associations or political parties, so honestly, part of me kind of wants it to pass đ€Łđ€Łđ€Ąđ€Ąđ€Ą
đš Il crĂ©e un script qui vide le budget du Parti socialiste via sa propre taxe sur les tĂ©lĂ©chargements ?
Un internaute a gĂ©nĂ©rĂ© 30⏠de taxe simulĂ©e en 4 heures en automatisant le tĂ©lĂ©chargement dâun fichier PDF hĂ©bergĂ© sur le site du Parti socialiste.
La dĂ©monstration intervient aprĂšs lâadoption Ă lâAssemblĂ©e nationale dâun amendement du dĂ©putĂ© socialiste Laurent Lhardit, instaurant une taxe de 0,055 ⏠par Go de donnĂ©es transmis aux abonnĂ©s français.
Avec un script simulant des tĂ©lĂ©chargements rĂ©pĂ©tĂ©s dâun PDF de 44 Mo consacrĂ© Ă Marc Bloch, lâingĂ©nieur a calculĂ© une taxe thĂ©orique de plus de 30⏠aprĂšs moins de 4 heures, pour 13 109 tĂ©lĂ©chargements et 576 Go de donnĂ©es. Il a ensuite prĂ©cisĂ© nâavoir effectuĂ© quâun seul tĂ©lĂ©chargement rĂ©el, les boucles Ă©tant simulĂ©es.
Il souligne que le texte nâexclut pas les associations ni les partis politiques. Selon son calcul, environ 1 000 personnes exĂ©cutant le mĂȘme script pourraient thĂ©oriquement Ă©puiser en moins de cinq mois le budget de 16 millions dâeuros du Parti socialiste.
Alex Agut retweeted
Human-Television interaction sucks so bad.
No more, you can now connect a simple microcontroller to your raspberry pi which runs a codex harness to control your TV
Alex Agut retweeted
Weâve raised $1.3M from YC & Afore Capital so you never have to type or speak again.
Typing is slow. Talking into your computer is awkward. Subvocal fixes both
Our glasses give you the speed of speech without you making a sound or touching a keyboard.
It tracks your internal speech muscles with non-contact sensing and decodes silent, almost invisible mouthing into text at more than 200 words/min.
Pre-orders are now live at:
subvocal.co/preorder?utm_souâŠ