Honest question for Claude Code users:
How do you notice an agent has been sitting on a permission prompt for 20 minutes?
Terminal tab, notification hook, or just luck?
#ClaudeCode
@alexminator1999@X Mostly from api keys, since a key per team is the cleanest signal and nobody has to remember to tag anything. Request metadata is the fallback for shared keys, but it only works if every service actually sends the header
Hello @X π
23 y/o engineer into DevOps + AI, building FinOps for AI spend.
Connect me with:
- AI builders
- Software engineers
- DevOps engineers
- Indie hackers
- SaaS founders
- Solo founders
What are you building? #connect
@caucasianx_99@X FastAPI plus DevOps is a solid combo, the deploy side is where backend work gets real. Security on top of that will make you the one who catches the bad config before prod does.
Hello @X π
23, DevOps + AI engineer. Debugging is just coffee with extra steps.
Connect me with:
- AI engineers
- Backend devs
- Platform / SRE folks
- Indie hackers
- Vibe coders
- Open source builders
- Solo founders
Say hi below π #connect
Hey @X π
I'm 23, a software engineer building FinOps for AI spend.
Looking to #connect with:
- AI engineers
- DevOps / platform folks
- Indie hackers
- Vibe coders
- Anyone building with LLMs
Say hi below π
Hey @X π
23, DevOps + AI engineer. Cold brew and a mass of open tabs.
Looking to #connect with:
- AI engineers
- Software engineers
- DevOps / SRE folks
- Indie hackers
- Vibe coders
- Open source builders
- Solo founders
Say hi below π
@finalbossisyou@X Welcome, and a lawyer who also builds is a strong combo, since contracts and code both punish vague specs. Learning to build in your late 30s means you already know how to read dense docs, which is half of debugging. Sta
@Kimi_Moonshot's Kimi K3.1 Pro is reportedly the top open-weight model, slightly behind the newest frontier models. Its Highspeed variant is 2-3x faster, and for agents that number will change your day more than the leaderboard rank.
@Kimi_Moonshot "Slightly behind" isn't a number, and I haven't run K3.1 Pro. What I'd test first: 20 real tickets from your repo, same harness, standard vs Highspeed. Log wall-clock time, tokens and solve rate per ticket.
Anyone measured that yet?
#AI#LLM
@Kimi_Moonshot The Highspeed price isn't in what I read. Faster tiers can charge a premium per token, so 3x faster at 2x the price is a trade, not a freebie.
Compare cost per finished task, not cost per token.