Thank you, and happy to answer questions on that, it's been a crazy time!
Maybe of relevance to non-security people here:
1. Most of it is about AI investigating event data in general, not just SOC/IR: cyber, intel, fraud, SRE, and we're even messing with customer 360 & social media data
2. For anyone into vibes coding or building agents, I encourage jumping to the "self-writing AI" section where we're finding we are moving internally from vibes coding -> vibes engineering -> and finally now to eval-driven AI coding loops
And, for anyone in security, doing careful evals here has indeed strongly colored my view on the market :)
Hey, I just saw your talk and for someone who's not really up to date with the latest AI developments it's eye opening what you got going in SoC investigations.
I personally work as pentester and we're still doing a lot of manual work with AI simply as a better version of Google, but seeing the BOTS presentation I feel we can do better. Do you have any idea if anyone's working on something similar to Louie in pentesting space, or if Louie could work with pentesting workflows?
Companies like xbow and horizon are using agents that talk to symbolic tools to automate more red teaming flows for different domains, so very much so. As shown in my talk, modern models are quite capable, and they aren't doing investigation-level scenario depth, more like scans, so seems like becoming the new expectation that everyone can & will do.
Companies like trail of bits are more interesting to me here, because they historically do deeper analysis. A place to look there is the darpa cc x ai (?) competition that finished at blackhat last year.
If in the US, we may be looking for a pen testing partner on an upcoming agentic AI contract, so feel free to msg - Leo @ graphistry
You are not a recovering audiophile at all, you are still fully in that rabbithole. Instead of enjoying music you ramble about your neighbor shaking and isolation etc.
You also didn't pivoted into preservation, it just happened because of whatever 'audiophile' thinking you think you have.
At the end you just stream music as everyone else.
The cover might gotten the relevant input from the band but the vinyl itself has only a handful of options and plenty of them are just black.
But come on do you listen to music of a band becase they are great in taking pictures? Or because they are really really good in standing next to a vinyl press?
No
You appreiciate their music and you don't need a commercial token to do so.
Naturally since people are buying things, technically they are consuming.
I mean that collecting a relatively small number of durable and visually pleasing objects isn't really the worst flavor of consumerism, even if it seems pointless to some people.
I agree we have a massive problem with over-consumption (most glaringly with things like fast fashion), but I'm not sure record collectors are a big problem.
There is still the law suite about FSD and the old hardware.
There is still Elon the hitler Musk Oligarch who wielded a chainsaw.
There were plenty of FSD videos last year and the year before showing that FSD is working. The question is still, is it working good enough, and what will be the business of a robotaxi.
The Taxi market overall is not that big, competition is hard and the most critical thing is peak demand.
In parallel random people believe tesla will wipe out the whole taxi industry + private cars tomorrow. Ignoring competition and everything else.
Aaand as an edit: When it finally works, people will tell you "told you so look at it, FSD works" yeah really? Of course it works but it was promised from Musk that 2020 all these Teslas will drive autonomsly. Its 2026
The performance of V3 and V4 have improved significantly. I use them daily, more V3 than V4.
It's partially about fully automated cars, but that's barely started. IMO, it's more about them as ADAS now.
And it's not just about whether anyone else will catch up in terms of automation/ADAS, it's about whether anyone else will catch up in terms of manufacturing+automation/ADAS.
Edit- And yes, Elon acting like that doesn't help, but Tesla isn't Elon.
Elon is Tesla. He owns 10-15% of it and the last vote for his trillion payout thing shows what this means.
And there has never been a person like Elon, interfering in german election, USA election etc. and being the posterchild for companies.
In what timeframe did V3 and V4 improved for you? At least for me, even FSD 2023 videos were quite impressive.
Nvidia has its own platform with ML based training. Im pretty sure car companies can and will just use them if necessary. Besides that, left and right other companies are working on it, i saw an xpeng driving autonoms through the city and it worked very well.
I think his decisions have been helpful in the past, especially compared to half-hearted attempts at EVs by most of most other large manufacturers, but only time will tell if his current positions will pay off. At least the new compensation is performance based. If he can't deliver, he doesn't get paid.
V3/V4 have significantly improved in the past few months. I use one, the other, or both, daily, more V3 than V4. They are well ahead of where they were in 2023.
I'm sure Nvidia and others will eventually catch up, but they have to catch up in terms of auto manufacturing/use, inference/training/sensor hardware manufacturing/use, fleet training data, etc... simulataneously.
Xpeng driving autonomously is great, and shows they are catching up at least in a specific situation. They may be catching up in the aggregate, they may not be. I agree that a Chinese manufacturer has the best chance to catch up overall.
In general, I think it's about whether a distributed, lower-compute/sensor with higher-data/training approach like Tesla/etc have will beat a more singular, higher-compute/sensor with lower-data/training like Waymo/etc have.
Distributed has been able to offer better ADAS at a lower cost, but singular is winning at fully autonomous driving. If distributed can catch up in terms of autonomous diving while continuing to improve ADAS, distributed is done. I have a hunch that's a part of why there's so much demand for DRAM/etc...
you take Elon out of Tesla and you will end up with Ford, $13-16/share
> At least the new compensation is performance based. If he can't deliver, he doesn't get paid.
Unfortunately, with Elon and TSLA, this is not the case. He just has to promise he will deliver in some imaginary future (as he’s been doing for more than a decade) and he’ll be handsomely rewarded
Market cap he’ll hit by President’s Day after few tweets about some amazing thing that is coming “by June” (year won’t be specified) and operational goals he’ll fudge with some “orders” from his buddies which will never materialize
Elon ain't magic. The first tranche alone requires another $1.2+ trillion in market cap.
He might have been able to push prices up to +/-10+% when Tesla was smaller, but I doubt he could even move it 5% these days.
Have they faked their current sales? How are they going to fake future sales?
It's clear you're not a fan of Elon or Tesla, but they'll have plenty of opportunities to mess things up themselves like any other company. There's no need to speculate about trillions in fraud and millions of fake sales.
If they hit the targets, he gets equity, if they don't, he doesn't.
I don't believe the West can sustain its lead. The US became rich by being the worlds manufacturing powerhouse in 1900. Over time, they grew comfortable, developed a massive upper class, and pivoted toward service and 'knowledge' jobs. They outsourced production to China or simply allowed China to take it over.
Germany and the EU followed a similar path. We know how to build machines and industrial products; precision and detail are our stereotypes and trademark. However, we also grew comfortable, focusing on services and high-level strategy while we did not invest to fix energy prices, raw resource dependencies and labor/automatisation.
How long is China looking fo resources in Afrika? Despite China being huge and having plenty of?
China overtook Germany as the leader in industrial exports in 2018, and now in 2026, the gap is already to broad. China has mastered machine building, controls the entire supply chain, and possesses modern technology; all while maintaining lower labor costs and a massive workforce.
Even if the USA and Germany try to avoid Chinese products, the rest of the world will not. We are entering a new era: the mass production of affordable, precise machinery globally, powered by China. If regions like Africa buy their solar, wind, and batteries from China, their entire energy grid and the machines running on it will be Chinese. They will look to China, not the West.
For example, a German company making weaving machines recently noted their price must be €60k to stay profitable, while a Chinese machine of the same quality costs only €20k. Once technology reaches its peak, differentiation is no longer possible to justify higher prices. The car industry is next; cars are becoming a commodity with shrinking profit margins. This shift will make China incredibly wealthy over the next 20 years.
And it wasn't even out of the blue. The shit was written on the wall and despite that what happens? We in germany discuss bureaucracy, if we should change our energy grid, IF investments in cheaper energy is reasonable etc.
We can’t even build our own infrastructure anymore. Look at the SuedOstLink disaster: while China builds massive 'Super Grids' in record time, Germany has spent over a decade and billions of Euros just arguing about a single power cable. Because we are stuck in bureaucracy and 'Not In My Backyard' protests, our energy costs are skyrocketing, while China doesn't care but still beats us in renewables.
It’s the perfect example of a society that has become so comfortable it has forgotten how to actually build the physical foundations of its own wealth.
And adding on top of all of that: AI and Robotic progress is fast, so crazy fast than when its here, we might have solved the other issues i mentioned...
Apparently we all have enough money to put it into OpenAI.
Some players have to play, like google, some players want to play like USA vs. China.
Besides that, chatting with an LLM is very very convincing. Normal non technical people can see what 'this thing' can already do and as long as the progress is continuing as fast as it currently is, its still a very easy to sell future.
I don't think you have the faintest clue of what you're talking about right now. Google authored the transformer architecture, the basis of every GPT model OpenAI has shipped. They aren't obligated to play any more than OpenAI is, they do it because they get results. The same cannot be said of OpenAI.
If Gemini can create or edit an image, chatgpt needs to be able to do this too. Who wants to copy&paste prompts between ai agents?
Also if you want to have more semantics, you add image, video and audio to your model. It gets smarter because of it.
OpenAI is also relevant bigger than antropic and is known as a generic 'helper'. Antropic probably saw the benefits of being more focused on developer which allows it to succeed longer in the game for the amount of money they have.
It'll just end up spreading itself too thin and be second or third best at everything.
The 500lb gorilla in the room is Google. They have endless money and maybe even more importantly they have endless hardware. OpenAI are going to have an increasingly hard time competing with them.
That Gemini 3 is crushing it right now isn't the problem. It's Gemini 4 or 5 that will likely leave them in the dust for the general use case, meanwhile specialist models will eat what remains of their lunch.
> Who wants to copy&paste prompts between ai agents?
An AI!
The specialist vs generalist debate is still open. And for complex problems, sure, having a model that runs on a small galaxy may be worth it. But for most tasks, a fleet of tailor-made smaller models being called on by an agent seems like a solidly-precedented (albeit not singularity-triggering) bet.
> But for most tasks, a fleet of tailor-made smaller models being called on by an agent seems like a solidly-precedented (albeit not singularity-triggering) bet.
not an expert by any means, but wouldn't smaller but highly refined models also output more reproducible results?
But then again the main selling point of using LLMs as part of some code that solves a certain business need is that you don't have to finetune a usecase-specific model (like in the mid 2010s), you just prompt engineer a bit and it often magically works.
>Also if you want to have more semantics, you add image, video and audio to your model. It gets smarter because of it.
I think you are confusing generation with analysis. As far I am aware your model does not need to be good at generating images to be able to decode an image.
It is, to first approximation, the same thing. The generative part of genAI is just running the analysis model in reverse.
Now there are all sorts of tricks to get the output of this to be good, and maybe they shouldn't be spending time and resources on this. But the core capability is shared.
I think you're partially right, but I don't think being an AI leader is the main motivation -- that's a side effect.
I think it's important to OpenAI to support as many use-cases as possible. Right now, the experience that most people have with ChatGPT is through small revenue individual accounts. Individual subscriptions with individual needs, but modest budgets.
The bigger money is in enterprise and corporate accounts. To land these accounts, OpenAI will need to provide coverage across as many use-cases as they can so that they can operate as a one-stop AI provider. If a company needs to use OpenAI for chat, Anthropic for coding, and Google for video, what's the point? If Google's chat and coding is "good enough" and you need to have video generation, then that company is going to go with Google for everything. For the end-game I think OpenAI is playing for, they will need to be competitive in all modalities of AI.
