Why is Gemini 2.5 (Pro and Flash) the only models willing to output more than 200 lines of code?
I asked multiple OpenAI models to create the rubik's cube simulation and they tried to do it with less than 200 lines of code.
It just doesn't work...
For the confused, it's actually super easy:
- GPT 4.5 is the new Claude 3.6 (aka 3.5)
- Claude 3.7 is the new o3-mini-high
- Claude Code is the new Cursor
- Grok is the new Perplexity
- o1 pro is the 'smartest', except for o3, which backs Deep Research
Obviously. Keep up.
City of Tokyo released a 3D digital twin of the entire city in high resolution point cloud, free to download. It’s an inevitable trend that more and more cities, houses, and factories will be transported into simulations.
Robots will not be trained in isolation. They will be simulated as an “iron fleet”, deployed in real-time graphics engines, and scaled across a huge cluster to produce the next trillions of high quality training tokens. The majority of embodied agents will be born in sim, and transferred zero-shot to our real world when they are ready. They will share a “hive mind” that sends latent embeddings back and forth to coordinate a multi-agent physical task.
Fun fact: NVIDIA’s Santa Clara Headquarter buildings are designed and rendered in Omniverse, a GPU-accelerated graphics platform, before materializing in atoms.
"Chief Automation Officer" - A scrappy semi-technical generalist
they've been 10x'd with LLMs, cursor, zapier ai, etc.
@stephsmithio called it out back in '23
who's doing this as a service?
As software becomes commoditized, taste - encompassing design, branding, user experience, and cultural relevance - is now the key differentiator and competitive advantage in Silicon Valley's tech industry.
https://t.co/0xL1nzePcn
Very cool, place well under “feel the AGI” category. As mentioned in the post, making actual apps is a lot more than code, you have to set up the entire environment, deploy it, etc. Automating all of this other infra will allow anyone to quickly build and deploy entire web apps.
In De-extinction tech has advanced, with plans to bring back species like the woolly mammoth, but it's uncertain if this will benefit ecosystems or just create more problems. https://t.co/gJVpkw3R09
I love reading books with @ChatGPTapp's new Advanced Voice Mode alpha.
It's a game-changer.
I sit on the couch with ChatGPT perched above me, and talk to it as a I read.
Here's what it gives me:
1. Instant clarification on complex ideas
2. Historical context at my fingertips
3. Up-to-date scientific insights
4. Chapter summaries to prep my brain
5. Opposing viewpoints for balanced understanding
It's like having a 24/7 expert companion, lowering the barrier to curiosity and deepening comprehension.
Sure, there are limitations—ChatGPT doesn't have direct access to the books themselves. But as the tech improves, this could revolutionize how we learn and engage with literature.
Imagine diving into Nietzsche or Plato with an AI tutor always ready to explain. Or dissecting scientific claims in real-time as you read.
The future of reading is here, and it's interactive, intelligent, and incredibly powerful.
Anthropic (@AnthropicAI) just announced a game-changer for their API: Prompt caching.
Think of prompt caching like this: You're at a coffee shop. The first time you visit, you need to tell the barista your whole order. But next time? Just say "the usual."
That's prompt caching. Here's why it's a big deal...
It makes it cheaper and faster to call any large prompt components that you reference over and over again.
Specifically? Up to 90% cost reduction. 85% faster responses. Use cases are wild.
Think:
•Cheap conversations with chatbots that have encyclopedic knowledge
•Coding assistants that can efficiently read your entire codebase
•AI that can discuss entire books at crazy low latencies and costs
It means a bunch of AI-powered apps that were previously too expensive to run just became viable.
Most companies suck at setting goals.
But Google, Amazon, and Microsoft figured it out ages ago.
Here’s their simple but powerful goal-setting framework (that you can use personally too):
Very interesting - In 2022, Dario Amodei chose safety over fame and fortune by withholding Claude, an advanced AI, from release, to prevent a potentially dangerous AI race. https://t.co/eqNSa58Gc4
@markgurman 13/ AI watermarking is impossible.
Marc Andreessen says detecting deepfakes will not work because AI is already too good, so the solution is to certify content as real.