You basiclaly have to be an Olympic-caliber skiier just to prepare the Olympics Alpine Sky track with a backpack paint-sprayer (blue lines extra dark for contrast because visibility hard AF when flying over 130km/hr with sun glare on white snow)
Average cost for 1 gigabyte of storage:
45 years ago: $438,000
40 years ago: $238,000
35 years ago: $48,720
30 years ago: $5,152
25 years ago: $455
20 years ago: $5
15 years ago: $0.55
10 years ago: $0.05
5 years ago: $0.03
Today: $0.01
We're hiring our first PM at Hello Patient and I think it's the most exciting role we've opened so far.
You'll work directly with me and alongside our Conversational Designers and Customer Engineers to ship AI agents into real healthcare environments.
This is the perfect role for an ex-founder turned PM or a highly technical PM who's owned complex products before.
My DMs are open and link to JD in next post
You can chat with your documents privately in Locally AI.
Everything is processed on-device and never leaves your device.
Powered by MLX for fast, private document understanding on Apple devices.
In a couple hours, we'll be rolling out About This Account globally, allowing you to see the country or region where an account is based. This will be accessible by tapping the signup date on profiles.
This is an important first step to securing the integrity of the global town square. We plan to provide many more ways for users to verify the authenticity of the content they see on X.
And for those in countries where speech has penalties, we've included privacy toggles to only show your region.
This was a huge undertaking. A special thanks to the engineers who made this happen: @sandeep_rao@singhai@mingsong@striedinger@rzhao0506
The security vulnerability we found in Perplexity’s Comet browser this summer is not an isolated issue.
Indirect prompt injections are a systemic problem facing Comet and other AI-powered browsers.
Today we’re publishing details on more security vulnerabilities we uncovered.
So, reminder: the quality of code output by these systems is *very low* and the AIs themselves don't understand the output.
This is obvious to anyone who knows how to program.
There are still use cases, for example, to output a large volume of low-quality code that is not intended to be used for serious purposes and is not expected to have a long lifetime.
Anyone extoling the virtues of AI code generation, who actually knows how to program, will do so with the context of the above paragraph in mind.
Anyone running around saying "the AI just generates all this code and it's great!!" either:
(a) Does not understand code and should not be trusted with making decisions about code;
(b) If representing themselves as people who understand code, are obvious frauds, or else have so much Dunning-Kruger they don't know how bad they are.
This may improve in the future. I would love it if I could have an AI help me write complex programs more quickly. But it's just not the state of things today, and anyone claiming it is, is either lying or being fooled.
https://t.co/H4jCDVnm8V
If you use Chrome and have upgraded to macOS Tahoe, disable "experimental prediction for scroll events" at chrome://flags to make it not feel like you just downgraded to a RAM-starved Windows 98 PC.
(Thank you, @jespervega!)
Michelangelo’s Pietà recreated with drones above St. Peter’s Basilica in Vatican City, as part of the ‘Grace for the World’ celebration on 13th September.
🔺iPhone models announced today include Memory Integrity Enforcement, the culmination of an unprecedented design and engineering effort that we believe represents the most significant upgrade to memory safety in the history of consumer operating systems. https://t.co/ule9gaXzc1
Remember when I wrote this weekend that people don't understand the unit economics of AI yet and that agents doing constant work and or running 24x7 will not cost the price of a Netflix subscription?
That's because they can't. Anthropic and OpenAI lose money on every $200 subscription, let alone $20.
It's loss leader pricing at a level we have never seen. About 3000%-4000% off the actual cost of running these models.
This poor fellow is experiencing it now. Everyone will soon.
The solution many folks present is "clear the context window" aka make the model stupider by giving it amnesia. One developer was optimizing his sleep schedule around Anthropic limits.
There is no solution currently.
This pricing is not sustainable at all with current technology.
The actual answer is that any agent that is not an open source model, running on you own hardware, will likely be out of reach for many folks in the not too distant future without breakthroughs in small model architectures or a complimentary breakthrough in large model architecture/distilling/etc, and that agents will primarily be digital workers for businesses and priced accordingly, aka at roughly human salaries.