Eric Levine

@piersonmarks @garrytan @chrysb Yes, this is the way. Implemented that in Clawvisor two days ago and documented in the SKILL.md.

@garrytan @chrysb @ericlevine Every API should have a POST /feedback so agents can leave issues when they face something that didn't work as expected. talk to your users (agents)

This is the coolest fun thing that I’ve been doing with my open source friends @chrysb and @ericlevine We just send each other bug reports from our claws trying to get things done and hitting snags It’s very helpful GitHub meets Moltbook should just have “prompt reports”

@garrytan @chrysb And it has been insanely valuable to get these bug reports / feature requests. When your power users are moving at 100x speed, you gotta hustle to keep up.

@nicenicenice122 @garrytan "Hello, this is Eric's personal assistant, Clawby. How can I help you?" After that we just had a nice conversation. Hadn't heard their voice before, so it really humanized my agent.

@ericlevine @garrytan well, what did it say?

I dropped @garrytan's gbrain Twilio skill into my OpenClaw, and it one-shotted it. In just a few minutes, it literally called my phone and started talking to me. Holy shit.

@garrytan @garrytan This is insane. I just dropped the link to my OpenClaw and it did the rest, one shotted it and in a few minutes it called my phone and was talking to me.

Here's the actual GBrain Skillpack Recipe github.com/garrytan/gbrai… This is one of my recent realizations: Just-in-Time software means YOUR Claw/Hermes will custom code something that works for YOU and your needs. This is Personal AI software, not packaged software.

I'm having a Her moment over here with my OpenClaw. It's literally the craziest thing in the world to be able to just talk to your claw and have fully customizable voice. Did you ever want to improve your ChatGPT Voice experience? Now you can have it totally under your control

AI coding tools have made building products much harder. In the past, 90% of your time writing code was easy or even boilerplate. 10% of the time you'd run into an issue that required hard thinking. Now, the AI does all the easy parts, and you're left with only the hard parts.

@jeiting @paulg CTOs pulling all nighters building “throwaway” MVPs are the original vibe coders. But in seriousness, I’m not surprised by the data, but imo it’s actually not that much harder to build while actually supervising the coding agents.

@ericlevine @paulg I think they will get some PMF then do a rewrite post-validation. Too early to tell

Vibe coding is here to stay. I'd been worried it might be a fad, but I talked to the founder of an infrastructure company who's in a position to see how well vibe-coded apps are doing, and he said a lot of them are making money.

@paulg I don’t disagree with the data at all, but I think companies that are built on vibe code will be outcompeted by companies where the engineers are in the details because they can ship features faster and scale when necessary.

@sudheerkumarme @IBIJBPM @iamshraey_ @akyshnik @Deivy_m_r I completely agree. If there was an easy way to find exceptional candidates from non-elite backgrounds when you have thousands of resumes to filter through, any business would be thrilled. It’s one of the reasons I’m bullish on @nextbyte_ai. Any other ideas?

The hard part isn’t just finding talent — it’s recognizing it beyond polished resumes, elite colleges, and algorithm-friendly stats. Some of us are honest, hungry to build, and capable of shipping great things, but often invisible in traditional filters. We just need a fair shot.

Soham also used to guilt trip me for his being slow on PRs when India Pakistan thing was going on, all while he was in Mumbai. Next person should hire him for the Chief Intelligence Officer role.

@sporadica For real. Many people apparently misunderstand the term “world-class talent”. Needles in haystacks.

So many mfs on twitter today acting like hiring isn’t hard you’re a wagie and have never had to run a company, stfu

@IBIJBPM @iamshraey_ @akyshnik @Deivy_m_r I think you underestimate how hard it is to find very talented engineers.

@iamshraey_ @akyshnik @Deivy_m_r Why did you offer to relocate him on a visa instead of just hiring a stateside American for the same salary?

@akyshnik That was fast! (Since we talked)

LETS GOOOOOOOOOOOOO 🦅🦅🌭🗣️🇺🇸🤑💪

Every founder has thousands of bad pitches in them and the only way to get rid of them is to talk to customers.

Shocked at how bad the UX is on the Uber Eats app. The restaurant we wanted to order from is exclusively there, but to order without joining their subscription service was several nonobvious clicks out of the store menu. No wonder DoorDash is winning.

Claude Code is the fastest, most self-assured, completely naive and (often) idiotic junior coder I've ever worked with.

@garrytan They know you already have the muscle memory to create a new meeting, invite yourself to it, copy the link and paste it into slack. What could be simpler

Zoom is so strange — they went to all the trouble of doing a calendar integration and the one thing I need is to flip an offline meeting to a Zoom meeting and they gray it out so I can't get a link and update the event directly from Zoom

@patrickc @garrytan Though the argument is weakened by the lack of data from the mid-2010s, my recollection is the tea party really kickstarted the wave of anti-intellectualism, but it didn’t really get into high gear until the pandemic.

@patrickc @garrytan If you take distrust in scientists as a proxy for anti-intellectualism, it seems to fit the data, though I would expect it to be a leading indicator of a drop in reasoning.

FT takes up the topic: "Have humans passed peak brain power?" ft.com/content/a8016c…

It turns out being able to debug vibecode is still important.

LLMs, and especially the code-writing AI tools, are absolute expert gaslighters.