carrys

[carrys @ccarrrys]

Anthropic CEO Dario Amodei on Claude:

"Claude's creators are building increasingly powerful AI while confronting one terrifying question: how do you control intelligence that keeps getting smarter?"

here's what this reveals about the AI race:

→ scaling AI could push intelligence beyond human capabilities - and accelerate technological progress

→ the challenge isn't just building more powerful models, but understanding and controlling their behavior

→ the companies that solve AI alignment could shape the next era of artificial intelligence

bookmark this ↓

[carrys @ccarrrys]

Anthropic CEO Dario Amodei on Claude:

"Claude's creators are building increasingly powerful AI while confronting one terrifying question: how do you control intelligence that keeps getting smarter?"

here's what this reveals about the AI race:

→ scaling AI could push intelligence beyond human capabilities - and accelerate technological progress

→ the challenge isn't just building more powerful models, but understanding and controlling their behavior

→ the companies that solve AI alignment could shape the next era of artificial intelligence

bookmark this ↓

[carrys @ccarrrys]

Claude engineers just dropped a 90-minute workshop on agents that check their own work

one agent executes
another verifies
if it fails, it tries again

13:51 → build your first agent
31:18 → add skills and specialist agents
34:20 → connect shared memory
35:23 → set outcomes and verification rules
53:57 → schedule recurring work
55:48 → “dreaming”: learn from past sessions

90 minutes
the full agent-building workflow

save this before you build another agent that just says “done” ↓

[carrys @ccarrrys]

@Serantych wait there’s an imessage one?? that’s the first one i’m checking lol

[carrys @ccarrrys]

@Serantych sheesh, didn’t realize GrokBot could handle this much. gotta try some of these, thanks!

[carrys @ccarrrys]

Claude engineers just dropped a 90-minute workshop on agents that check their own work

one agent executes
another verifies
if it fails, it tries again

13:51 → build your first agent
31:18 → add skills and specialist agents
34:20 → connect shared memory
35:23 → set outcomes and verification rules
53:57 → schedule recurring work
55:48 → “dreaming”: learn from past sessions

90 minutes
the full agent-building workflow

save this before you build another agent that just says “done” ↓

[carrys @ccarrrys]

Claude engineers just dropped a 90-minute workshop on agents that check their own work

one agent executes
another verifies
if it fails, it tries again

13:51 → build your first agent
31:18 → add skills and specialist agents
34:20 → connect shared memory
35:23 → set outcomes and verification rules
53:57 → schedule recurring work
55:48 → “dreaming”: learn from past sessions

90 minutes
the full agent-building workflow

save this before you build another agent that just says “done” ↓

[carrys @ccarrrys]

This is f*cking insane. SpaceXAI engineer Jadiah, ex-Cursor:

"99% of people use GrokBot like a chatbot. I use it as an autonomous colleague that coordinates worker agents, manages tasks, and gets real work done while I sleep"

in this 18-minute talk, he explains how GrokBot goes beyond simple chats

GrokBot → Coordinator Bots → Worker Agents → Autonomous Workflows

one agent coordinates the work. others execute tasks in parallel

smarter models are only half the equation. the system behind them is what makes autonomous work possible

watch and bookmark it today ↓

[carrys @ccarrrys]

@choopyplug1 wait, one security bug was hiding in plain sight for 21 years? the 3-pass method sounds worth trying

[carrys @ccarrrys]

This is f*cking insane. SpaceXAI engineer Jadiah, ex-Cursor:

"99% of people use GrokBot like a chatbot. I use it as an autonomous colleague that coordinates worker agents, manages tasks, and gets real work done while I sleep"

in this 18-minute talk, he explains how GrokBot goes beyond simple chats

GrokBot → Coordinator Bots → Worker Agents → Autonomous Workflows

one agent coordinates the work. others execute tasks in parallel

smarter models are only half the equation. the system behind them is what makes autonomous work possible

watch and bookmark it today ↓

[carrys @ccarrrys]

@vartekxx wait, you can actually do this with Opus 5.5 now? that’s crazy

[carrys @ccarrrys]

This is f*cking insane. SpaceXAI engineer Jadiah, ex-Cursor:

"99% of people use GrokBot like a chatbot. I use it as an autonomous colleague that coordinates worker agents, manages tasks, and gets real work done while I sleep"

in this 18-minute talk, he explains how GrokBot goes beyond simple chats

GrokBot → Coordinator Bots → Worker Agents → Autonomous Workflows

one agent coordinates the work. others execute tasks in parallel

smarter models are only half the equation. the system behind them is what makes autonomous work possible

watch and bookmark it today ↓

[carrys @ccarrrys]

you’re not learning. you’re collecting bookmarks

Claude can turn those saved posts into:

→ workflows you can steal
→ prompts you can reuse
→ tutorials you can follow
→ ideas you can actually build on

24 practical examples
5 categories
one copy-paste prompt

stop confusing saving with doing

save this before you bookmark another post ↓

[carrys @ccarrrys]

@Serantych try the same task before and after adding this
count how many times you have to step in

[carrys @ccarrrys]

you’re not learning. you’re collecting bookmarks

Claude can turn those saved posts into:

→ workflows you can steal
→ prompts you can reuse
→ tutorials you can follow
→ ideas you can actually build on

24 practical examples
5 categories
one copy-paste prompt

stop confusing saving with doing

save this before you bookmark another post ↓

[carrys @ccarrrys]

@Serantych give it a finish line and clear boundaries otherwise you’re just approving every footstep

[carrys @ccarrrys]

you’re not learning. you’re collecting bookmarks

Claude can turn those saved posts into:

→ workflows you can steal
→ prompts you can reuse
→ tutorials you can follow
→ ideas you can actually build on

24 practical examples
5 categories
one copy-paste prompt

stop confusing saving with doing

save this before you bookmark another post ↓

[sunick @Serantych]

GrokBot + REA is f*cking dangerous...

that feature you keep paying for?
start by finding out how it actually works

trace the code
uncover the rules
build your own version

here’s the full workflow, with prompts you can steal ↓ https://t.co/BTFCOFX6Vs