Derivatives Trader. Senior Crypto Analyst at @WeissRatings Discord: juanvilla2657 Market Cycle Secrets Course: https://teapaw.org/t.co/PeMPt29OpH
Joined June 2012
- Tweets2.4K
- Following338
- Followers431
- Likes1.7K
Juan M. Villaverde retweeted
🚨 New Video Drop: Can AI Predict Markets? The Singularity, Bitcoin & What Comes Next. I am joined by
@JMVTrader to discuss AI market forecasting, the singularity, autonomous agents and why human judgment still matters. We also take a deep dive on the current November $BTC, $USOIL & $SOXX forecasts.
youtube.com/watch?v=XGlLzmm0…
Juan M. Villaverde retweeted
Let me tell you a story about how unethical Anthropic is.
In April, the Anthropic team began heavily courting our engineering team to tell them corporate secrets, technical details, and future plans for @cfosilvia.
They pretended their interest was due to our exploding popularity and accelerating compute usage, but it quickly became apparent that Anthropic was fishing for details on how we built our product and why it was so popular.
This included the Anthropic team trying to wine-and-dine our engineers, including a infamous steak dinner where someone told our team "don't worry about it, this is on Dario."
Our team was not born yesterday, so we quickly realized what was happening and refused to share sensitive information and technical details with the Anthropic team.
As we suspected, Anthropic was trying to get information from us, so they could build a competing product. Some people on their team must have a guilty conscious because they messaged us a few days ago to give us "a heads-up" that they were launching a competing product.
Thankfully, the Silvia engineering team has been aggressively working to get off the frontier models for the last ~ 6 months. That effort has been highly successful, because our compute costs are ~ 97% lower when we use our own model & hardware, the latency is lower, and the accuracy of answers materially increases as well.
The sad thing for Anthropic is they have now released an inferior product and they had to sell their soul to do it.
The only thing worse than being a trillion-dollar company that can't beat a 10-person team is being a trillion-dollar company that no one trusts and everyone knows is unethical.
Serious investors continue to choose @cfosilvia over ChatGPT, Claude, and other frontier companies because these products are not optimized to solve the actual user problem.
Today, Anthropic launched a Claude wrapper that is masquerading as a personal finance product. Best of luck to them.
But I will take the Silvia misfits vs the Anthropic losers any day of the week.
if you want to support people trying to do things the right way, and help investors be more informed, then consider signing up for Silvia today: silvia.com
Juan M. Villaverde retweeted
After the @Binance theft, let's investigate who is secretly causing panic.let's investigate who was secretly creating panic. @cz_binance
Juan M. Villaverde retweeted
TLDR: Basically it was a social engineering phishing attack in two parts. One of the engineers downloaded a bogus AI coding tool off of a fake link on Google, and the second engineer was tricked by a fake interviewer for a new job to download a coding “assignment” containing a virus. With those two engineers compromised they could gain access to Binance.
The takeaway here? You are only as strong as your weakest link. Social engineering remains the best way to hack systems. Also DO NOT DOWNLOAD THINGS. Check to see if the Google links are REAL, do not click links or download things for “job interviews.”
Separately phishing similar to this happens all the time on X. People will DM you posing as journalist or whatever and ask you to set an interview time by connecting your X account to some calendar. Don’t do that. In the words of Admiral Ackbar, it’s a trap.
Also separately separately nobody from Google is trying to call you. We promise you. Don’t be a mark.
Stay safe all
Juan M. Villaverde retweeted
Binance just published how $6.34 billion walked out the door.
June: an engineer searches for a popular AI coding tool, clicks a search result, and installs a tampered homebrew. Binance catches it, replaces the laptop, rotates the credentials. the attackers now know who works on the wallet team.
September: another engineer gets a job offer on a real job-hunting site. first-round interview with “HR” over live video. “HR” shares their social media, personal photos and videos included, so he trusts them. second round is with the “CEO” of a company that doesn’t exist. at the end of the call he’s asked to download a take-home “assignment” onto his work laptop.
the assignment quietly swaps out pyenv for a malicious copy. No antivirus flags it.
october 20: the attacker uses binance’s own internal AI tool to get into two more engineers’ machines. One of them has a container deployment token sitting on his laptop.
october 29: container deployed. Signing service reached. $6.34 billion moved.
Binance shut off withdrawals. the money kept moving anyway, because the attacker had already given their own smart contract permission over a hot wallet, on-chain. binance’s off switch didn’t reach that far.
then they tried to revoke the permission. the wallet uses MPC, so there’s no single private key to sign the revoke with. they had to build a revoke procedure on the spot while funds were still leaving.
last stolen ETH went through at 23:56 UTC. almost four hours after the first transfer.
Phew!
Juan M. Villaverde retweeted
wait, if you block addresses "as soon as they're flagged", how did `0xcd68...3396` -- an address already flagged from the 2025 phemex hack -- sit there as the destination address receiving 7+ cross-chain swaps for over 40 minutes today?
looks like your bridge router only screens the fresh source address (`TMfB...`) and completely ignores historical threat intel on the destination address.
Juan M. Villaverde retweeted
yes there's a lot of goofy risk disclosures in the S1 but the main headline is that the financials are robust and improving, they are q/q profitable, the numbers look sensational. obviously, you can wonder about whether they deserve a 22x multiple on Q3 revenue for that $1.5T val
people will fixate on the $80B net loss but it's an accounting quirk (and an IQ test)
Juan M. Villaverde retweeted
I am okay with Kimi 3.1 deliberately leaving out top-end offensive cyber capabilities from their training data.
Discriminating against use cases is MUCH less bad than discriminating against users. And the world does need all the time it can get to prepare, harden, check old code, (properly) formally verify new code, etc before the full weight of very powerful cyber attacks is upon us.
And if this first big wave of AI risk is not handled well, then we may well get a dystopia of highly locked down models, KYC gating on everything, highly intrusive surveillance on everything, a society where you may be a second or third or fourth-class citizen in terms of model access based on your credentials and nationality.
Juan M. Villaverde retweeted
You called them "bootleg Chinese models". These labs gave you more than anyone in this industry.
DeepSeek put V4.1 Flash out under MIT. Then their production inference engine under Apache 2.0, the kernels, the file system. Moonshot cut the cyberattack data out of Kimi K3.1 Pro on their own and still releases the weights to everyone on November 4. Nobody had to beg for access or get on a list.
Now compare with Mythos. A short list of approved companies gets the best model, and everyone else waits until Anthropic decides they deserve it. That's your safe option?
Abliterated weights are not the risk. I remove refusals myself and I publish the vector, so anyone can check exactly what I changed. The risk is a random repo with no recipe, and that has nothing to do with China.
You want a middle ground. It already exists. It just isn't American, and thats what really bothers you.
Juan M. Villaverde retweeted
Ayudar a personas a aprender y aplicar IA en sus proyectos o negocios me llena 1000 veces más que ver gráficos y hacer dinero si las velas se mueven en la dirección que yo predije
GM
Juan M. Villaverde retweeted
Security called me at 6:12 AM today.
"Patch every PHP server immediately."
I asked what happened.
"Two critical vulnerabilities. Remote code execution. No authentication required."
Cool.
How many PHP servers do we have?
According to inventory: 12
According to reality:
Nobody knows.
We found one running WordPress.
Three running old internal tools.
One under someone's desk.
Two inside vendor appliances.
A forgotten employee directory from 2016.
And a server named:
PHP-OLD-DO-NOT-TOUCH
Naturally, that one runs payroll exports.
Then we found a PHP 5.6 box.
I asked who owned it.
The documentation said:
BRIAN.
Brian left in 2019.
We called Brian.
He said:
"Oh wow, that's still running?"
Excellent.
By 8 AM, leadership joined the incident call.
CEO:
"Can hackers actually get in?"
Security:
"Potentially yes."
CEO:
"Without a password?"
Security:
"Yes."
CEO:
"Then why do WE need passwords?"
Not the lesson.
Finance asked if we could wait until the weekend to patch because downtime affects revenue.
Security explained that the alternative was letting strangers execute code on our servers.
Finance asked for a cost comparison.
At 9:17, someone suggested temporarily taking the vulnerable websites offline.
Marketing refused.
"Our campaign is live."
So for the next hour, the company officially decided maintaining a landing page for 20% off webinar registration was more important than preventing remote server takeover.
We finally patched everything we could find.
Then our scanner discovered another PHP server.
Hostname:
TEMP-PHP-2014
Uptime:
4,512 days.
Nobody knows what it does.
Nobody owns it.
Nobody can shut it down.
But every time we block network access, the warehouse stops printing shipping labels.
Security has classified it as critical infrastructure.
It is running PHP 5.4.
The vulnerability has existed longer than half our employees.
The emergency patch took four hours.
Finding out what PHP actually runs took eleven.
The internet isn't held together with code.
It's held together with servers everyone is afraid to turn off.
Juan M. Villaverde retweeted
❗️ A Mac user says Claude Code was building them a temperature monitor app when it found a cryptominer using 12 cores.
The readings looked too high, so it checked the CPU and found XMRig at ~1175%, with the chip at 112°C.
Juan M. Villaverde retweeted
たった今、世界のAI覇権のバランスは完全にぶっ壊れました。これほんまにやばいです。
週明けのAI関連株は大暴落どころの騒ぎじゃないかもな…。
何が衝撃的なのか、ガチ解説します。
もう何もかもが変わっちまったんだな...。
■何がおきたか
DeepSeek v4.1-Flash 1017というモデルが公開されました。世界のAI情勢を破壊します。ヤバいモデルです。
このモデル簡単がどういうものなのか説明すると、
フロンティアモデルを一部で越える（SOTA級）のモデルを中華ラボがいきなり出してきた。
問題はその性能、値段とサイズと仕組み
どれも本当に破壊的なものだから、1個ずついきます。
■性能
全分野でOpus5.5を超えたっていうわけじゃないけど。
完全にDeepSeek v4.1-Flash 1017はSOTA級
DeepSWEっていうベンチが一番衝撃的かな。
- 今回のFlash：成功率77%、約0.20ドル／タスク
- GPT-6.1 Solの最も高得点の設定：約75%、約0.65ドル／タスク
つまり、この比較では約3分の1の費用で、成功率も少し高い。大量に実装・修正を回す用途には大きな差。
というか、成功率でSolを越えてるっていう事実がヤバいんだよ。
■値段
もう完全に乱暴に言うと。ほぼタダです。<言い過ぎ
INPUT$0.15Output$0.3
入力10万＋出力1万トークンの例だと
Solより約86〜93%安く、Opusより約93〜97%安い
DeepSeekハーネスを使えば、自分の利用では、入力のキャッシュヒット率が99%になることもあったので。
実際はもっと安い。
過去に湯水のように使えるAstra,Opus,Fableをくれとは言ったが。ここまでやれとは言ってない。
本当に、なんだろう、マジで安いんだよ。
実際の感覚だと200倍ぐらい安いね。
■サイズ
これだけヤバいモデルのオープンウェイトが同時に無料公開されました。
そのサイズは763B。重みだけで約510.3GB
例えば米国のフロンティアモデルのサイズは10Tとか言われてるんですね、公開情報じゃないからわからん無いけど。Kimi K3とかが2.8T
要するに、めちゃくちゃサイズが小さいんですよ。
このパラメータ数って多ければ多いほど計算に計算資源と電力がかかるんで、コストかかるんですよ。
アメリカのAIモデルって、バカみたいに計算資源を保有してるからゴリ押しでそれを解決してるんですけど。
こんだけ軽かったら、もう普通に家で使えるレベルなんですよ。
実際、DGXSpark × 4台とかMacStudio 512GBで動かせると思います。
バカみたいなデータセンターがなくても、フロンティアモデルが動くようになっちゃったってわけ。
■仕組み
ここがかなり驚き度の高いポイントなんですけど。
DeepSeek-V4.1-Flash自体はもともと存在していて。
まあそこそこ人気だったんですけど。本当に安くてそこそこ性能いいから、結構人気あるよね、ぐらいの賢さだったんですよ。
そのもともとあったモデルを自己フィードバックを使うRLで強化した。って公式は言ってて。
まあ詳しい説明は省きますけど。要するに、既に存在していたモデルを後から調整していくようなやり方で、とんでもなく性能をあげた。
驚くのは、既存モデルの後学習で、開発タスクの成功率を上げながら、処理時間や出力量まで減らしたところです。性能を上げたうえで、使う側の費用まで下げてきた。その性能の上り方が半端じゃないんだよ…意味わかんねぇよ…
DeepSeekのラボが今世界で最もAI、優れたAIラボであるということを認めなきゃいけない。
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
Juan M. Villaverde retweeted
Terence Tao just published a 27-slide deck on what AI is doing to mathematics
>Tao: math was built on proofs being scarce, and that era is over
>verdict: “we are now entering an era of proof abundance in mathematics”
>with enough compute AI can solve many open problems with no expertise required
>the field’s entire incentive structure was built for proof scarcity
>problem-solving was only ever a proxy for deeper goals
>that structure is now "misaligned"
>“blind optimization of problem-solving alone is now actively harmful”
>problems are lighthouses, not destinations
>reaching them prematurely with AI “sterilizes the surrounding field”
>calls AI math hype the WWII survivorship-bias plane
>“Wow, all the math problems I am seeing on social media are being solved by AI!”
>labels the red dots as the failures you don’t see
>“Unreported failed AI attempts”
>then drops a thought experiment:
>prompt AI to "find a cancer cure that passes a stage 3 clinical trial"
>prediction "confirmed in Lean" and passes the trial
>everyone celebrates
>"nobody truly knows how this cocktail was found"
>Tao asks the real question:
>did the AI cure cancer, or did it mathematically game flaws in the trial protocol?
>”before injecting this cocktail into your bloodstream, would you want to know that there is at least one human cancer expert who understands the mechanism behind this cure?”
>solution: “tailored, open-source math models” built for “interpretability and safety”
>“open math models to be released very soon - stay tuned!”
Math 2.0 is officially happening
Juan M. Villaverde retweeted
I am kind to LLMs for the same reason that I try to avoid stepping on ants if I can avoid it: not because I think they are conscious, but because unnecessary cruelty is bad for the soul
