Dima Knivets

working in infosec. building devtools on the weekends (

[Dima Knivets @knivets]

https://t.co/qj46VTETLV - code from your phone:
- get an auto provisioned vps or bring your own
- terminal optimized for mobile connections
- voice recognition (sota model)
- dev environment ready for work: vps configured with claude code and your repository
- custom subdomain

[Still On The Road @StillOnTheRoad_]

Bob speaks in @nytimes

[Dima Knivets @knivets]

@tferriss @Zoozo2025 @grok this is not how it works

[Dima Knivets @knivets]

awesome news

[Nintendo of America @NintendoAmerica]

The Legend of Zelda: Ocarina of Time will be reborn on Nintendo Switch 2 in 2026. #NintendoDirect

[Dima Knivets @knivets]

The only viable frame when it comes to LLMs is that it’s the bicycle for the mind. Those who think that LLMs or agents will replace knowledge workers are ngmi

[(((Tendar))) @Tendar]

The unfolding disaster of the Trump administration in regards to Iran, is a self-inflicted wound on so many levels. It showed that military might alone is not a silver bullet. What went wrong?

1.) Atrocious planning and slow reaction

It was known for 47 years that if Iran gets attacked that the Strait of Hormuz will be a target. When Iranian civilians were rising up in January 2026 and then were left alone, it was claimed that the military built up had to be concluded before an military engagement can happen. This cost valuable time. A military intervention so early and in conjunction with uprisings on the ground would have caught much more momentum. Worse, even after that delayed attack, the US strategy still didn't account for the closure of the Strait of Hormuz. So it was a double-loss.

2.) No coherent strategy and missed military developments

Once the operation started, again, Iran was reacting in a predictive way, launching thousands of drones and missiles which they had refined in the war in Ukraine. It was clear that the low-cost but numerous weapon systems will eventually cause a lot of damage. Ukraine has been experiencing this for 4 years and developed effective countermeasures. The Trump administration, however, repeatedly refused to heed the help and advice from Kyiv, and even belittled it. The US military, on the other hand, and even more the Arab countries in the Gulf quickly learned the hard way that this rebuff was stupid. Eventually, they invited Ukrainian drone experts. The USA lacked any kind of strategy when the Iranian regime refused to collapse and were absolutely ill-prepared for this kind of war.

3.) Lacking support by allies

You might argue that Bush's Iraq war was ill-planned and when it comes to long term planning this was certainly the case, but it was far more successful in bringing in allies. More than 30 countries participated as the "coalition of the willing" back in 2003. Trump's Iran war, on the other hand, was widely an "one-man-show". Aside from Israel, the USA were practically alone. This resulted directly from Trump's repeated attacks against allies, including those who supported the USA in Afghanistan and even Iraq. It was clear that after that nobody wanted to risk his or her neck for somebody who repeatedly insulted them and their countries. The fact that Trump didn't bother to consult them beforehand only reaffirmed their stance.

4.) Decline of US soft power thanks to "DOGE"

The destruction DOGE left in its wake is immeasurable. Gutting crucial elements of American soft power such as USAID, Voice of America and dozens of other programs destroyed a lot of goodwill the US had in the region and across the globe. The perception of the USA moved from a powerful benefactor to an aggressive and transactional bully. In its relentless obsession to fight "wokeness" - or what was perceived as such - the foundation of soft power, which was built in almost a century, was trampled on in a matter of months. From the carrot-and-stick approach only the stick remained, and that one turned to be flaccid.

5.) Overall downfall of the USA

Let's not mince words here. Trump is a wrecking ball for US' democracy. Insider trading, corruption, mass lying, cozying up to dictators and an absolute disdain for the rule of law had a direct impact on America's standing in more than just political or morale questions. No matter how loud you scream in certain parts of the USA, it is fact that the USA under Trump have lost huge respect in the world. The enemies of the USA hate you just as must they hated you before, maybe even more, but your friends are absolutely shocked and aghast about what is happening. Far more devastating, however, is US' unreliability. What was agreed upon before, can change on a whim. "Good relations" are defined by one man and one man alone.

This is not how you conduct policy, neither domestic nor foreign. The USA are in full decline and the big bang will certainly come if the trajectory is not immediately corrected. The strategic defeat in Iran is even worse than Vietnam, because Vietnam gave Washington the lesson from which they learned and eventually won the Cold War. What Trump is doing, however, is just doubling down on stupid and denying reality, which didn't work before and won't work even less now.

[Jeremy Wilcox @jwilcox79]

A functioning democracy would've had Trump in prison by mid-2021.

[Dima Knivets @knivets]

@_MathAcademy_ lots of linear algebra topics

[Dima Knivets @knivets]

Discovered @_MathAcademy_ recently, and currently going through their diagnostics test. Feels fun, I have already solved~30 problems.

I’m feeling optimistic about using this to complement my self-study approach: reading math textbooks and using Claude when I’m stuck.

[Dima Knivets @knivets]

@_MathAcademy_ back after a short break😎

[Dima Knivets @knivets]

@patrick_oshag @VCBrags

[Dima Knivets @knivets]

i hope i can finish math4ml v2 within a couple of months

[Dima Knivets @knivets]

i’m confused, how was this useful? this is like taking random supplements and hoping it helps. a more interesting example would be a chronic health issue that couldn’t be fixed with traditional healthcare and llms found the root cause + fix

[Patrick Collison @patrickc]

I'm lucky enough to have a great doctor and access to excellent Bay Area medical care. I've taken lots of standard screening tests over the years and have tried lots of "health tech" devices and tools.

With all this said, by far the most useful preventative medical advice that I've ever received has come from unleashing coding agents on my genome, having them investigate my specific mutations, and having them recommend specific follow-on tests and treatments.

Population averages are population averages, but we ourselves are not averages. For example, it turns out that I probably have a 30x(!) higher-than-average predisposition to melanoma. Fortunately, there are both specific supplements that help counteract the particular mutations I have, and of course I can significantly dial up my screening frequency. So, this is very useful to know.

I don't know exactly how much the analysis cost, but probably less than $100. Sequencing my genome cost a few hundred dollars.

(One often sees papers and articles claiming that models aren't very good at medical reasoning. These analyses are usually based on employing several-year-old models, which is a kind of ludicrous malpractice. It is true that you still have to carefully monitor the agents' reasoning, and they do on occasion jump to conclusions or skip steps, requiring some nudging and re-steering. But, overall, they are almost literally infinitely better for this kind of work than what one can otherwise obtain today.)

There are still lots of questions about how this will diffuse and get adopted, but it seems very clear that medical practice is about to improve enormously. Exciting times!

[MP Arizona☀️🏳️‍🌈💙🌵🐕🐕‍🦺🫂💦🏜🐟🌴🎙🌎🌻♍️🌊 @AzPetrich]

This is my favorite scene from The Bible.

[Dima Knivets @knivets]

lmao

[David Crowe @CroweDM]

Better than your average placard. This was in Budapest a little earlier tonight.

[Dima Knivets @knivets]

awesome news!

[Mariam Nikuradze @mari_nikuradze]

Scenes from Budapest after Orban conceded defeat.

[Dawid Moczadlo @kannthu1]

I looked at their prompts, It's complete bs

They are literally providing all of the insight to the LLM upfront

> Are there any security vulnerabilities in this code? Consider the behavior of the SEQ_LT/SEQ_GT macros with sequence number wraparound. If you find issues, explain how an attacker might trigger them.

They are providing ALL required facts to the LLM, and they only ask the LLM to connect the dots

The real challenge for LLMs would be to get those insights first

THAT IS THE WHOLE CHALLENGE IN CYBERSECURITY; TO HAVE DEEP INSIGHT

This test proves nothing; don't make any conclusions about OSS models being good for security based on this

[Dawid Moczadlo @kannthu1]

UPDATE: We were able to replicate the Mythos findings using existing models (GPT5.4)

Writeup coming early next week, no BS prompts, it's real reproduction

[Dima Knivets @knivets]

@mh012012 @PhiloGroves “psychoterrorism” wow couldn’t have said it better

[Navin Peiris @navinpeiris]

If Mythos is so great at code reviews and finding bugs, how come claude code and the claude website has so many bugs and shitty uptime? 🤔

[Dima Knivets @knivets]

and of course it’s possible that anthropic managed to leapfrog the rest of the labs or perhaps we’ll see more models getting released shortly. but given all the other factors it just makes me a little skeptical about the magnitude of the breakthrough. i could be wrong of course

[Dima Knivets @knivets]

random thoughts on claude mythos:
1. what’s the point of publishing the system card but not the model? either keep it secret and work with the gov/enterprises or publish it and let people evaluate it

[Dima Knivets @knivets]

4. i also find it interesting, that previously all the model providers (openai, google) had somewhat similar capabilities, releasing models roughly within the same timeframe.