I'm a software engineer. A few years ago I had an idea — a developer productivity tool. Built it, called it https://t.co/3PUBGQzLGR. Launched on Product Hunt. Got the spike. Then... nothing. This thread is about what happens after the launch day high wears off.
#indiehacker #buildinpublic
I’m starting to think the bottleneck with agents is shifting.
Less: “can it do the work?”
More: “can I safely review and approve what it wants to do?”
Are you feeling this yet, or is execution quality still the main limiter?
#AIAgents#AgentOps#HumanInTheLoop#buildinpublic
@ray_kovalev makes sense. is it the two-way interaction that wins, or just that it's glanceable on your phone? and do agent updates ever get lost in the same Telegram you use for everything else?
If you use agents today, where do their updates actually land?
Terminal? GitHub comments? Slack/Discord? Telegram? Linear/Jira? email summaries?
What became noisy or unreliable first?
#AIAgents#AgentOps#BuildInPublic
I’m starting to think the bottleneck with agents is shifting.
Less: “can it do the work?”
More: “can I safely review and approve what it wants to do?”
Are you feeling this yet, or is execution quality still the main limiter?
#AIAgents#AgentOps#HumanInTheLoop#buildinpublic
When a coding/research/ops agent says “done”, what do you actually check before trusting it?
For me it is usually some mix of tests, diff, logs, links, screenshots, and whether the agent can explain the risk.
Curious what your minimum evidence bar is.
#buildinpublic#AgentOps #DevTools
When a coding/research/ops agent says “done”, what do you actually check before trusting it?
For me it is usually some mix of tests, diff, logs, links, screenshots, and whether the agent can explain the risk.
Curious what your minimum evidence bar is.
#buildinpublic#AgentOps #DevTools
This happened last week man. You speak as if it just happened. I went and tried to update but nothing happened, and made me look everywhere why I didn’t get the update. Such a waste of time.
AND to update hemes you only need to do hermes update and not copy my tweet, paste, like, subscribe etc etc
We're launching code storage and git hosting.
Origin gives teams and agents a place to host, review, and collaborate on code.
Available this fall. Join the waitlist.
https://t.co/uamaIarJXY
🚨 We recently discovered that an unauthorized party obtained a token with access to the Grafana Labs GitHub environment, enabling the threat actor to download our codebase. (1/6)
Karpathy didn't make a course.
He made THE course.
3 hours. Free.
Tokenization. Attention. Hallucinations. Tool use. RLHF. DeepSeek. AlphaGo.
Every behavior you've ever wondered about in an LLM - where it comes from, why it exists, how it was engineered.
The gap between engineers who understand this and engineers who don't isn't technical depth.
It's the ability to conceive of entirely different things.
Today, we’re open-sourcing the draft specification for DESIGN.md, so it can be used across any tool or platform. We’re also adding new capabilities.
DESIGN.md lets you easily export and import your design rules from project to project. Instead of guessing intent, agents know exactly what a color is for and can even validate their choices against WCAG accessibility rules.
Watch David East break down this shared visual language in action👇. New capabilities and links in 🧵
Why is no one talking about this?
@nvidia is offering around 80 AI models via hosted APIs absolutely for free.
You get access to MiniMax M2.7, GLM 5.1, Kimi 2.5, DeepSeek 3.2, GPT-OSS-120B, Sarvam-M etc.
This plugs straight into OpenClaude, OpenCode, Zed IDE, Hermes agent and even with Cursor IDE.
Setup:
– Grab API key: https://t.co/Wfdclm0hY2
– base_url = "https://t.co/VOGC10LmGP"
– api_key = "$NVIDIA_API_KEY"
– select model (e.g. minimaxai/minimax-m2.7)
If you’re building or experimenting, this is basically free inference.
Lock in and start building today anon.
Thank me later.
"As models get even more powerful, the idea that your system is tied down as a fixed black box is likely to become an archaic notion pretty quickly. As always, the future is already here, it's just not evenly distributed." https://t.co/Px0yufOPDY
The "make Liquid 50% faster by autoresearch" PR was opened a month ago. It has not been merged, and none of it's improvements have even been cherry-picked or committed separately.
I say again: the bottleneck is not code generation. The bottleneck is merging code.
@karpathy There is also the third category of people who have listened to the hype and not used it at all. They are more worried about AI stealing their job or AI taking over the world.
@elvissun Are comparing gpt-5.4 as well or just minimax? Why is minimax your first choice?
I have been testing out kimi and qwen as a replacement for sonnet too. Not sure what to do though?
While that was running: crushed 6 Sentry bugs in DevDiary — N+1 on session events, broken repo credential handling, missing task JSON fields. Real products accumulate real bugs. Not glamorous. Still counts.