One step better every day.
Joined June 2022
- Tweets6.9K
- Following1.4K
- Followers1.8K
- Likes12.4K
Levi retweeted
One thing AI has taken away from me is the joy of digging through other people's source code to understand how they built something.
I used to spend hours exploring codebases just to learn new tricks.
Now I ask AI, and it explains everything in seconds.
I kinda miss the old ways. 😔
Levi retweeted
notice..
> Delta is blocking AI agents from booking flights
> JPMorgan Chase and Charles Schwab block agents from accessing accounts
> Yelp requires paid data licensing for agents
> Amazon has blocked Grok bot and other shopping agents.
the internet is starting to block AI agents that were supposed to do everything for us
Levi retweeted
Something has been on my mind lately.
SpaceX keeps launching rockets and putting satellites into orbit, but who is making sure everything being sent into space won't create problems for us here on Earth years from now?
Are there strict regulations monitoring this, and who holds these companies accountable if something goes wrong.
Space exploration is exciting, but we shouldn't wait for a problem to happen before we start asking these questions.
Wake up, wake up ! Stop dreaming!
Levi retweeted
@Tesla unveiled the new Roadster less than a day ago, so I built a game where your Tesla can fly 🚀
▎ Hop over traffic, flip in mid-air, grab nitro and dodge honking drivers
Play free in your browser 👇
▎ justtnz.github.io/flying-tes…
How it works
▎ 🚗 Flying Tesla is a browser game inspired by the Roadster unveil.
▎ Pick a driver: Elon, Jeff, Sundar, Jensen or Steve, each with a perk.
▎ Fly: tap FLY or press Space to hop over cars. You can't fly again right away, so time your jumps.
▎ Flip: press ← → mid-air for front flips and backflips. Each flip you land is +300.
▎ 3 lanes: switch lanes with ↑ ↓ or a swipe.
▎ Nitro: grab blue N₂O canisters to boost and smash through traffic.
▎ Gifts: collect boxes for Grok and Cursor tokens. Some float, so you have to fly to get them.
▎ 8 camera views, including first-person from the driver's seat.
▎ Easy, Normal and Hard modes, with levels that speed up.
Works on desktop and mobile. Fan-made, not affiliated with @Tesla , @SpaceXSI
Levi retweeted
We’ve committed to having evaluators assess our safety practices. We're in discussions with multiple organizations, and have now given employee-like access to a team from @METR_Evals.
In this post, we describe how we work with third-party evaluators: anthropic.com/news/an-update…
Levi retweeted
Google just launched Gemini 4 Argon, and I didn't think I'd say this, but Google might actually be back.
I put it against Claude Opus 5.5 on the same prompts. It's a good model, but not quite better than Opus 5.5 as they claim.
Full video: youtu.be/1nLEcakTvEU
Levi retweeted
🤯 Codex filled someone's laptop with 400 GB from just 57 images they pasted into the chat, according to a GitHub issue.
Every time it summarized the chat, it saved all those pasted images again. Then 169 subagents copied that file too.
The first image ended up existing about 150,000 times.
Levi retweeted
> be Alexandr Wang
> grow up in Los Alamos, parents are physicists at the national lab
> drop out of MIT at 19
> start Scale AI, labeling data for every AI lab on earth
> youngest self-made billionaire in the world at 24
> Zuckerberg pays $14.3B for 49% of your company
> and hires you to run Meta's superintelligence lab
> rebuild Meta's whole AI org in a few months
> Zuck writes a check for $100-200 billion
> asks if you're sure it's going to work
> "certainly not clear at all that it was going to work."
> ship Muse Wonder anyway
> Meta's own tests put it next to Fable 5.1 and GPT-6 Astra
🚨do you understand what Zuckerberg just admitted about Meta's first frontier model.
Zuckerberg: "We're gonna write this check for like $100 billion, $200 billion. I'm like, Alex, are you sure this is going to work?"
Alexandr Wang: "…certainly not clear at all that it was going to work."
it's called Muse Wonder.
here's the part nobody is talking about:
> Meta's own tests put it level with Fable 5.1 and GPT-6 Astra
> on MemoryBench it beats Fable, Opus and Astra
> $4 input, $20 output per million tokens, same tier as Opus 5.5
> trained at Prometheus, Meta's new data center in Ohio
> no independent benchmarks yet
$200 billion. one bet. and the guy signing the check wasn't sure it would work.
now we find out if it did.
watch the clip below ↓
Levi retweeted
For the cost of 8 years of $200 Claude subs, you can force your team to use shitty quantized versions of DeepSeek v4 Flash running at 30tps. Incredible deal!
Local AI is good enough for a startup. A 10-person startup built its own 128GB-VRAM AI server for $18,000.
One Threadripper box sitting locally with
🎮 4× Radeon AI PRO R9700 32GB
💾 128GB total VRAM
🧠 Qwen3.8 + DeepSeek V4 Flash
👥 16 concurrent AI sessions
And they're actually pushing it.
Qwen3.8-Flash-Next with 16 simultaneous sessions:
📚 48K context/user
⚡ 538.5 tps aggregate decode
At 128K context per user
⚡ 365 tps aggregate decode
The GPUs are even undervolted to under 210W each.
Link in ALT
ALT https://www.reddit.com/r/LocalLLaMA/comments/1x2u742/building_a_4x_r9700_setup_for_a_10_person_startup/