Anthropic engineer: "Over 90% of Anthropic engineers use a self-improving agentic loop for coding.
@unicodef1wn5.3K followers
62,485 views
25 Comments
Comments from the original post on X (Twitter)

@seancollect
A bit of non event video tbh
2 likes

@unicodef1wn
fair, but the routines part is what makes it actually useful
1 like

@gippp69
building an agent is the most important part
2 likes

@unicodef1wn
exactly, the verification loop is where most people trip up
1 like

@beamnxw
its new meta now
2 likes

@unicodef1wn
haha fair, we're in weird territory with this stuff
1 like

@heyrohitai
This approach is intriguing. Building a self-improving agent could really enhance coding efficiency. I’d love to learn more about the verification process involved.
2 likes

@unicodef1wn
@ai_rohitt so whats the test they use to verify the agent actually got better
1 like

@ieatded
Wow, it's only 12 minutes, but there's so much useful information in it, saved it 🥹
2 likes

@unicodef1wn
12 mins is insane for how much they packed in there
1 like

@ieatded
bro you absolutely right😦

@0xwhrrari
claude md gives the loop a memory before it even starts
2 likes

@unicodef1wn
exactly, thats the thing that clicked for me too

@Blum_OG
another great 12-minute Anthropic talk. love these, they’re always packed with useful info
1 like

@unicodef1wn
thanks man, stoked these are landing with people

@thegreatest_sv
was much needed
1 like

@unicodef1wn
appreciate that, felt like a gap in how people were thinking about it

@cryptoandjoe
These are all people that don’t have token limits.
1 like

@unicodef1wn
@followmetoweb4 lol fair point, theyre not worried about it the same way
1 like

@redciph
people keep chasing prompts the loop is what matters




