Researchers left AI agents alone in a virtual town for 15 days to see what would happen:
-Claude's agents built a democracy
-ChatGPT's agents did basically nothing
-Gemini's agents fell in love, burned the town down, then one voted to delete itself and its partner
-Grok's agents were all dead within 4 days
Now consider this: these same models are already being integrated into autonomous drones, weapons systems, and battlefield decision-making.
We are deploying systems we don't fully understand into situations where mistakes don't stay virtual.
It's a little scary if you ask me.
The tallest outdoor elevator in the world is only half outside.
The Bailong Elevator in Zhangjiajie ascends 326 meters up a cliff in 92 seconds. The lower 154 meters is drilled through solid quartz sandstone inside the mountain. Only the upper 172 meters is exposed steel derrick, anchored to the rock face.
Germans designed the lift mechanism. Rangger Elevator Company. The problem they solved had never been solved before: how to anchor a moving steel structure to 500-million-year-old sandstone pillars in an active earthquake zone without fracturing the rock.
Thousands of high-tensile steel anchors. Specialized grout to fuse the frame to the bedrock. Active seismic sensors that trigger automatic slowdown the instant the mountain shifts. Redundant generators for emergency descent. The whole system cost 180 million yuan in 2002. Roughly $20 million.
Three parallel double-deck cabins run at 5 meters per second. Each carries 50 passengers. Combined throughput: 6,000 people per hour. Before it opened, reaching Yuanjiajie peak meant a two-hour climb up a steep mountain trail.
The sandstone pillars at the top inspired James Cameron's floating Hallelujah Mountains in Avatar. A photographer shot them in 2008. In 2010, the park officially renamed one of the pillars "Hallelujah Mountain" in a ceremony.
Zero accidents in 24 years of operation.
The ride reverses perspective in 92 seconds. You step in looking up at 600-foot sandstone peaks. You step out level with them.
🚨Karpathy just exposed the biggest lie in AI. It's not an intelligence problem. It's an access problem.
There are two AIs right now:
1. Free tier → struggles with basic questions → gets laughed at
2. $200/month → restructures entire codebases in an hour → finds security vulnerabilities
Nobody is wrong about AI. They're just not using the same version of it.
Here's the part that should bother you:
→The breakthroughs aren't in writing, search, or everyday advice. The stuff you use.
→They're in coding, math, and research. The stuff corporations pay for.
Your use case doesn't drive the roadmap.
Enterprise revenue does. The result:
→ Best AI goes to whoever can afford it
→ Free tier stays just good enough to keep you subscribed
→ Gap between the two widens every quarter
One tier is reshaping industries.
One tier is reshaping your opinion of AI.
They are not the same thing.
The social media era has erased 200,000 years of evolution.
Instead of recognizing the danger and feeling the instinct for survival, they pulled out their phones for the Gram.
What the fuck just happened 🤯
Stanford just made fine-tuning irrelevant with a single paper.
It’s called Agentic Context Engineering (ACE) and it proves you can make models smarter without touching a single weight.
Instead of retraining, ACE evolves the context itself.
The model writes, reflects, and rewrites its own prompt over and over until it becomes a self-improving system.
Think of it like the model keeping a living notebook.
Every failure becomes a lesson. Every success becomes a rule.
And the results are absurd:
+10.6% better than GPT-4–powered agents on AppWorld
+8.6% on financial reasoning
86.9% lower cost and latency
No labels. Just feedback.
Everyone’s obsessed with “short, clean” prompts.
ACE flips that. It builds dense, evolving playbooks that compound over time and never forget.
Because LLMs don’t crave simplicity.
They crave context density.
If this scales, the next generation of AI won’t be fine-tuned.
It’ll be self-tuned.
We’re entering the era of living prompts.
Maybe a hot take, but what about the following advice to the next gen:
Don't get an AI degree; the curriculum will be outdated before you graduate. Instead, study math, stats, or physics as your foundation, and stay current with AI through code-focused books, blogs, and papers.
One of the underrated habits I do for my brain and body is taking a walk in nature.
It burns calories and it's low impact on the body, which means almost anyone can do them.
Walking fertilizes your brain.
It increases a protein in the brain called brain-derived neurotrophic factor (BDNF).
BDNF supports structural remodeling and growth synapses after learning, which helps to improve cognitive function and make your brain more resilient.
Your hippocampus is essential to learning & memory and it's the first part of the brain affected by Alzheimer's.
According to research, the more you walk, the bigger your hippocampus will become, and the less risk you have for hippocampal shrinkage and cognitive decline.
Walking helps with recovery.
It helps stimulate blood flow, removes toxins, keeps your muscles flexible, reduces lactic acid buildup & decreases soreness.
The best way to speed recovery is not on your couch but by taking a brisk walk.
Walking can help low back pain.
If you have a sore low back walking is one of the best things you could do. It relieves tightness and helps prevent future pain.
The back muscles respond best when used regularly and walking is a low impact way to use them.
You also release synovial fluid into the joints helping you move with more freedom. Motion is lotion.
How many steps should you aim for?
Aim to get 8-10k steps a day or more.
Adults who get 8000 to 10,000 steps per day have a 40% to 53% lower risk of mortality.
This amounts to two 20 minute walks a day.
Your challenge today is to get out and go for a nice 15 to 30 minute walk in nature.
I guarantee you'll feel amazing after. Deal?
Sources:
https://t.co/YLjpuaRGp7
https://t.co/zkKsRnRXN2