STL😎Ncat '21 👨🎓 Supporting others in building successful ventures 💯I believe we can achieve our goals together. Let’s talk about building a SaaS product
@solomonneas@sflorimm Not sure. My agents have access to set up even the infrastructure for GCP. Recently tested using cloudflare and same there. They have a really nice MCP server as well.
I often don’t share this kind of thing because it’s usually AI slop.
But this article about building a Chief of Staff with Claude Code is one of the best real examples of agentic systems I’ve seen.
AI NATIVE SLACK IS JUST SLACK
THE BLOAT IN SLACK IS BECAUSE OF PEOPLE, NOT PRODUCT
MIDWIT HIRES LIKE TO CREATE PROJECTS AND BUSYWORK BC THEY LITERALLY CANNOT PRODUCE VALHE
AI DOESNT SOLVE THIS
If this Karpathy interview doesn't pop the ai bubble,
nothing will.
10 brutal quotes:
1. LLMs don’t work yet
They don’t have enough intelligence, they’re not multimodal enough, they can’t use computers, and they don’t remember what you tell them.
They’re cognitively lacking. It’ll take about a decade to work through all of that.
2. When you boot them up, they always start from zero
They have no distillation phase, no process like sleep where what happened gets analyzed and written back into the weights.
3. What’s stored in their weights is only a hazy recollection of the internet
It's just a compressed blur of 15 trillion tokens squeezed into a few billion parameters. Their context window is just short-term working memory.
4. They’re good at imitation, terrible at going off the data manifold
Too much memory, not enough reasoning.
We need to strip away the memorized knowledge and keep the cognitive core: the algorithms, the magic of intelligence, problem-solving, strategy.
5. We’ve probably recreated a cortical tissue, pattern-learning and general, but we’re still missing the rest of the brain
No hippocampus for memory.
No amygdala for instincts.
No emotions or motivations.
6. They memorize perfectly but generalize poorly
If you give them random numbers, they can recite them back. No human can do that.
That’s the problem: humans forget just enough to be forced to find patterns.
7. Anything truly new, code that’s never been written before, ideas that have no template; they stumble
They’re still autocomplete engines with perfect recall and no understanding. Until we find that cognitive core, intelligence stripped of memory but full of reasoning, they’ll stay brilliant mimics, not minds.
@VictorTaelin I have same issues. except I do try to use multiple codex chats to implement different features. It’s great if I have a hour or two to watch and test what’s been implemented but if I need to walk away from my computer when I come back I am sometimes lost at what I was doing.
Google did it again!
First, they launched ADK, a fully open-source framework to build, orchestrate, evaluate, and deploy production-grade Agentic systems.
And now, they have made it even powerful!
Google ADK is now fully compatible with all three major AI protocols out there:
- MCP: To connect to external tools
- A2A: To connect to other agents
- AG-UI: To connect to users.
AG-UI is the newest addition, which is an open-source protocol that enables agents to collaborate with users.
They worked with the AG-UI team to build this.
It takes just two steps:
- Define the agent with ADK, with its tools, state, etc.
- Connect it to any React frontend using CopilotKit.
The AG-UI protocol gives you a bridge between a backend AI agent and a full-stack app.
And CopilotKit provides the building blocks you need to bring your agents into frontend applications.
Find the repo in the replies!