Los agentes de IA siempre escriben código de más.
Por ello, acaban de lanzar Ponytail.
Le mete en la cabeza la mentalidad del senior dev más vago (y eficiente) de la empresa.
Antes de generar cualquier cosa, sigue esta escalera:
- ¿Esto realmente necesita existir? (YAGNI)
- ¿Ya lo hace la stdlib?
- ¿Hay feature nativa de la plataforma?
- ¿Dependencia ya instalada?
- ¿Se puede hacer en 1 línea? → Solo entonces escribe el mínimo que funciona.
Resultados reales medidos:
• 80-94% menos código
• 3-6× más rápido
• 47-77% más barato
REPOOO👇
Ayer recibí los dos informes de la UDEF sobre Zapatero, como me imagino le ha pasado a media España.
He hecho como Rufián el otro día y apenas he dormido leyendo todo por aclararme.
Como los medios sacan noticia por noticia para ganar clicks… he decidido ordenar las ideas y hacer un hilo completo sobre el caso.
El informe 1907 (186 páginas) reconstruye la estructura: quién es quién en la red, qué sociedades controlan, por dónde se mueve el dinero, dónde acaba y qué papel juega cada implicado. Es la radiografía societaria y financiera.
El informe 1908 (158 páginas) reconstruye la cronología: ordena por fecha los mensajes de WhatsApp, correos electrónicos, llamadas y reuniones recuperados del móvil de Rodolfo Reyes y de los demás dispositivos intervenidos. Es el guion temporal del caso.
Uno te explica el cómo y el cuánto, el otro te explica el cuándo y el con quién.
Por eso, a lo largo del hilo, las capturas alternan entre los dos: cuando se cita un mensaje concreto (una frase, una hora, un día), la fuente es casi siempre el 1908; cuando se cita una conclusión policial, una titularidad societaria o un flujo de dinero, la fuente es el 1907.
Abrochaos los cinturones… que empezamos.🧵
My brother-in-law runs his own data center/GPU farm at home.
Currently makes him ~$3k/month profit. 30% solar powered. Aiming to get to 60% by end of this summer which will increase profits significantly (mostly down to adding batteries, not panels), then will buy more GPUs.
He sells the compute via some 3rd party service that takes a cut. Apparently a lot of ML/rendering workloads.
You can just do stuff!
The new fee rate in @Polymarket is just INSANE in low probability markets (1 to 30%) yo pay up to 7% COMISSION ON CRYPTO AND 3% ON SPORTS. You literally have better chances on a sportbook I have no idea who thought that was a good idea but it's UNUSABLE PLIS FIX
From a trading perspective what makes JS unique is their drive to solve/master 2nd and 3rd order problems to solve the original problem. The most profound example of this is Fixed Income ETFs. They saw the incoming tsunami of interest in the early 2010s. They figured that to crack this market they'd have to become ace MMs in FI products, especially bonds. They then went and created probably the most efficient bond/rates desk on the street. When Covid hit and bond ETFs started dislocating, they managed to pick up % points on bonds that trades in a 5bp range. Their dedication to the "process" gives them the confidence to put on risk that almost any other firm would consider nuts. This commitment to the process is almost unmatched in time horizon, resources committed and risk taking.
FYI JS is a non entity in low latency trading.
@PolymarketDevs Los de polymarket lo de la backward compatibility no lo practican es de coña. De las apis más cutres que he visto, para encontrar algunos mercados tienes que ir a más matches y ni siquiera aparece en la API.
If you are using @Polymarket don't deposit directly from @CoinExSpanish I have no idea why but they block your account because they say the funds come from a place mixed with Iranian sources. Withdrawal first to a hot wallet and then you can send you poly, hope it help somone!
Es que son super básicos por ejemplo en este habla de que el precio cambia por un oráculo cuando el polymarket solo depende de bid y asks. Son para gente que no sabe del tema está claro pero nose porque lo hacen lol
¿Por qué últimamente hay tantos vendiendo humo con polymarket?
Hilos super básicos sobre bots mágicos que imprimen dinero en el mercado up down o de este palo.
No entiendo que venden, no veo su modelo de negocio o si son de polymarket para atraer volumen. 🤷
Muy pesados.
I BUILT A BOT THAT SEES THE FUTURE 8 SECONDS AHEAD
Penalty, red card, goal - probability jumps 15-30 points instantly.
But Polymarket updates in 15-20 seconds, my bot in 8.
Live stadium data through Sportradar WebSocket - player coordinates, speed, xG, pressing, all in realtime, not from broadcast.
$12K/year, same data bookmakers buy.
ML model recalculates probability after every event. Lineup, form, fatigue, minute, score. 3 months of training, $2K on GPU.
Private RPC on Polygon - $200/mo, 3-5 seconds faster than public nodes.
Gas overpay for block priority - $800/mo at 8 matches per week.
Up to 8 matches in parallel. Champions League, EPL, La Liga. Each match - separate instance, separate wallet.
Total cost: $18K/year.
But I see the future 8 seconds ahead.
I don't know who will win, but I know exactly where the oracle falls behind.
third worldism is not about something being safer or cleaner to visit as a tourist: minsk, pyongyang etc are all clean and orderly cities
in moscow specifically its the corruption that runs through every aspect of life:
- infrastructure
- healthcare
- public safety
- law enforcement
- regulation
you will get arrested for having a beer in the street, but an ethnic gang that runs over your kid may get away with it because they are under some elite protection
also russia is not really capitalist anymore, everything is either absorbed by giants or forcefully closed, small and medium businesses are consistently choked
a corrupt, totalitarian and decentralized system is not preventative, anything goes as long the local operator can continue to steal money, until a disaster happens: a fire, a crash, a mass poisoning, etc
introducing Bluma.
the all-in-one platform for AI UGC.
we’re the first to de-edit videos - breaking them into scenes, captions, and elements automatically.
Bluma lets you create winning organic short-form and paid ads with our asset generator and node-based canvas that saves your creative workflows.
we allow you to clone winning formats, batch generate assets, and edit videos all in one place.
comment “ugc” for free credits and early access to @getBluma!
https://t.co/ffvMa9weIP
Breaking: OpenAI fired Leopold Aschenbrenner at 22. Three years later, he manages $5.5 billion.
And is quickly becoming one of the best AI investors out there
Today, we're launching a tracker to trade alongside his picks, automatically. Here's his full portfolio:
1. Power & Energy:
$BE — Bloom Energy (29%) → fuel cells powering data centers
$EQT — EQT Corp (4%) → largest US natural gas producer
$CRWV — CoreWeave (4%) → GPU cloud infrastructure for AI
$SEI — Solaris Energy (3%) → energy infrastructure
2. Bitcoin Miners:
$CORZ — Core Scientific (14%) → Bitcoin miner turned AI data center host
$IREN — Iris Energy (11%) → Bitcoin mining + AI cloud
$APLD — Applied Digital (9%) → AI data center infrastructure
$CIFR — Cipher Mining (5%) → Bitcoin mining / HPC
$RIOT — Riot Platforms (3%) → largest US Bitcoin miner
3. Semiconductors & AI Hardware:
$SNDK — Sandisk (8%) → memory/storage for AI workloads
$COHR — Coherent (3%) → laser/photonics for data centers
$TSEM — Tower Semiconductor (3%) → analog chip foundry
$INTC — Intel (2%) → US chip manufacturing bet
$LITE — Lumentum (2%) → optical networking for AI infra
You can now mirror the portfolio automatically on Autopilot.
Just connect your broker, choose his tracker, and you're good to go
Link below
Equity perps right now look a lot like crypto perps did in 2024.
Currently funding runs 11-31% annualized and liquidity is thin.
But crypto perps faced the same criticisms until the basis trade arrived.
Before BTC spot launched on Hyperliquid, funding averaged around 18% annualized.
Within months it compressed to around 9%,
That same flow deepened the order book and anchored price to the underlying.
Equity perps could be approaching the same inflection point.
Right, if you want a civil war, ethnic bloodbath, and lawless descent into warlord barbarism like Libya, just arm the 10% Kurd minority that inhabits a tiny part (darker brown area of the map) of the country, and which is part of a stateless ethnic population of 60 million that is loathed by every country in the region. You see how well that worked in Syria---its got rid of a pluralist dictator in favor of a head-chopper, who then set about to kill both Assad's Alawites and the Kurdish opposition, too. That is, the people at the CIA who have cooked up these exercises in human death and misery should all be fired. The whole CIA should be splintered into a thousand pieces, just like JFK said!
So there's this 24 year old who got fired from OpenAI named Leopold Aschenbrenner who raised $1B then grew it into $5.5B... IN ONE YEAR
He dumped his Nvidia position and now he's buying every Bitcoin mining company in America. But not because of Bitcoin...
Because they already have the two things every AI company is desperate for:
- Power grid access
- Permits that take years to get
It's like buying a bar that already has a liquor license instead of waiting 3 years for one
These companies are already pivoting from crypto to AI hosting. He's buying them before the market prices it in.
The smartest trade in AI right now is not about the chips, it's all about how you get them powered
Most people trading Polymarket have never heard of Sasha Stoikov.
Meanwhile, the bots taking their money are running his math on every single quote.
This is Stoikov at IAQF in 2023 - an 80-minute lecture on where market making meets market microstructure.
The actual theory from the person who built it.
He covers what textbooks skip: how optimal quotes degrade under real order flow, why inventory pressure isn't linear in practice, and what happens when your Poisson arrival assumptions meet an actual limit order book.
My full breakdown of the applied Polymarket stack: Avellaneda-Stoikov adapted for binary settlement, GLFT inventory bounds, Glosten-Milgrom adverse selection, VPIN kill switches - is in the article.
Start with Stoikov's lecture. Then read the thread.
I just published a data dump of full order book data from @Polymarket
The data is maximally granular. There is no filtering whatsoever. Every order book change and trade is saved. Across all markets
Updates are hourly. Each snapshot contains ~30M rows. Snapshots are downloaded as parquet files. Each file is approx. 500MB-1GB large.
The data dump is already 2B+ rows large and growing fast. But this is just part 1/3. Coming soon is a much bigger dump that also includes @Kalshi / @opinionlabsxyz / @trylimitless etc
I started collecting this data because I noticed I couldn't get it from Dome API. Their historical order book data was filtered limiting its usefulness. Also now with the acquisition there's a lot of uncertainty about whether they will continue operating
the cybersecurity industry is about to get completely disrupted.. 💀
someone just open-sourced a fully autonomous AI Red Team.
it’s called PentAGI. multiple AI agents that talk to each other to hack a target. zero human input.
I don't understand why people aren't using Claude Cowork to crush local SEO.
I CAN OUTRANK YOUR LOCAL BUSINESS IN 60 DAYS WITH JUST CLAUDE COWORK.
Here's how I would do it: