This was a fantastic read. I had no idea how much of Terminator 2 had to be invented from scratch. It's amazing to think that a lot of the tools and ideas that shaped modern VFX started with engineers just trying to solve one impossible problem after another.
Some films really do stand the test of time, I'm not really sure that contemporary CGI will really age as well.
This looks very cool. An AI that can listen and speak and handle tasks without breaking the flow of conversation would solve some big annoyances with current tools.
The concern is though as these get better will people struggle to distinguish these with real human connections?
Interesting that even Nasdaq-100 inclusion could not keep SpaceX shares up. Probably the expected boost from index-fund buying was already priced in.
I hope that investors are finally looking beyond the hype, the valuation was massively over in my opinion. It is hard not to be conspiratorial about it given all the bending over backwards they did for the listing.
AI is being pushed as a productivity revolution. Sure, it is very good. But, the costs of it are astronomical and manufacturers and customers are left paying the whooping bills, in this case from energy. Who is really benefiting from this revolution?
I don't think the issue AI note-taking per se. It is whether a conversation can still be candid and confidential once everyone knows it may become permanent, searchable, and shareable. Big tech companies don't exactly have a clean record on privacy.
Just saying no sounds simple enough. But social pressure makes that very difficult.
The normalisation of constant recording really has me worried, everything is so monitored now. Privacy just keep get more and more eroded for the sake of minor convenient features and products. At what point do we start deciding these trade off are no longer worth it? And is it even too late to make that decision?
The idea that its two lobes joined through a slow, gentle collision is really cool. It could hold clues of how planets first began forming billions of years ago. Really interesting object.
What stands out to me is how quickly developers switched once they perceived the model became good enough for agents.
It may not be the best model overall but at that price it doesn’t need to be. OpenAI and Anthropic need to get their act together to bring down their eye-watering prices down or they will end up having their dinner eaten by competitors.
This is exactly why AI summaries shouldn’t be trusted blindly, especially when safety concerns can be buried under hundreds of positive reviews.
It is a big problem that the uncertainty in the text produced by LLMs isn't surfaced to users. It is also a big problem that companies think that shoehorning AI summaries everywhere is a good idea.
Watched this when it first came out last year. I think it is the clearest explanation of quantum computing I’ve seen. It is especially good at tackling the myth about trying every answer at once.
3blue1brown videos are always absolutely top quality.
EV batteries may be lasting far longer than most buyers expected but are consumers really going to trust the data over the headlines? There has been such a propaganda attack on EVs that it will take a long time to overcome that.
Fascinating read. The bit I’m worried about is not AI writing a novel, it’s the possibility of culture drowning in endless polished, but average writing. To be fair we’ve been on a similar cultural trajectory without AI, everything feels samey.
Really enjoyed this framing of threat modelling as a way to make assumptions explicit and not just a compliance checklist. It was also quite amusing and sassy. Well done to the author, great piece!
The point that secure is meaningless without defining the adversary and assets is especially important.
One thing it doesn't tackle that I would like to know more about is how do teams keep these assumptions and threat models current as the system and its environment evolve? I think that is a massive challenge.
People aren’t mad about technology. They’re mad about being told after the fact that their town might lose water, power, quiet, or tax revenue to a project they didn't have a say in.
Build the future, sure. But don’t sneak it past the people who have to live next to it. If you have to sneak it like that, then maybe it is not worth having.
Get sick of all the shady behaviour and lying. These companies, owners, and CEOs need to be taken down a peg.
This is where prediction markets get messy. The moment a ranking has a payout attached to it, people start looking for ways to move the ranking. These markets are going to create more toxic incentives. Their boosters aren't going to acknowledge that.
First thing is first, this is really cool.
This feels like the right way to frame LLM-assisted engineering. AI can generate a shocking amount of code, but the actual value is in the review discipline, and tests around it. The browser Kubernetes angle is cool, but what I find more interesting is the workflow, and especially testing behaviour against k8s instead of just trusting “looks right.” I do wonder how many teams are already doing this level of verification for AI-written code. It might be the direction everyone goes in over the next few years.