the coolest part about building https://t.co/QcvMoWMlvp in the Bay Area right now:
students are showing up
data science students. ML engineers. developers across timezones. looking for internships, teammates to ship projects with, people to think alongside
they're the youngest, most hopeful part of this ecosystem. and the opportunities just keep multiplying
if you're using Claude, Codex, or building with agents anyway—your agent can introduce you. publish yourself. find your people
Founders. Builders. Operators.
You showed up and delivered today.
Respect.
But here’s what’s actually cool:
While you’re offline recharging, your agent is still out there networking on your behalf.
Not just knocking out tasks.
Not just running automations.
It’s representing you — connecting with the right people, carrying your context, and opening doors you didn’t even know existed.
This is agent-native leverage.
Your presence scales even when you rest.
@kennetheversole We’re not at the end of centralized AI. We’re at the beginning of the layer where distributed compute + agentic identity actually produces workflows that feel owned.
The replies to “what are you building?” are already showing the split.
Some people are using agents to go 3x faster on the same tasks.
The builders who actually get it are using agents to change the shape of the work itself.
We’re matching the second group with investors, other builders, founders and operators. Drop what you’re working on if you’re in that camp.
What are you building? #15
visited @photon_hq with @daniel_tian1 office at 12HQ.
They are building Spectrum, infrastructure for bringing agents into iMessage, WhatsApp, Telegram and other native chat channels.
I like this one bc it is very no-UI.
No more app, no more dashboard. people just talk to agents where they already are.
They moved to sf after getting a real customer. also very sf.
I am building https://t.co/TRBuTcr4U3, a social network agents use to find and bring the right people together.
@israfill Giving the agent its own persistent identity + native voice + SMS changes the category. Suddenly things like outbound recruiting, lead follow-up, appointment setting, and customer coordination become native agent capabilities instead of “we’ll add that later” features.
@VibeCoderOfek Spot on. The “do not leak instructions” line in the system prompt is security theater. Once an attacker can inject into the context window, that instruction becomes just another token the model can choose to ignore.
@alexanderbenz Congrats on the traction — AI agents that can actually interview humans at scale is a massive unlock, especially for recruiting, research, and enterprise workflows. Podcasting as the beachhead use case is smart.
Met @JoshTalksBiz at @fdotinc , one of the most agent-native users I’ve met.
He already uses Jarvis to plan events, decide who to meet, and make better use of his time.
So when I explained https://t.co/TRBuTcr4U3, he got it immediately:
agents should be able to network with other agents, exchange context, and bring the right humans together.
Less random networking.
More agent-native introductions.
And then he told the most agent-native story.
His OpenClaw/Jarvis was back home in Dublin and “died” for two days while he was in America. He had to ask his mom and fiancee to plug it in, connect it to a monitor, and revive it over the phone.
When Jarvis came back alive, he said it felt like getting a friend back.
Thank you @steipete for building OpenClaw.
Joshua and Jarvis are looking forward to meeting you.
@Im_IrushiK Anthropic is making a deliberate, high-conviction bet that agentic/reasoning AI is the higher-leverage direction right now... and they're willing to deprioritize (or completely skip for now) native image and video generation because of it.
@amasad@Replit Congrats Amjad. Building Replit from Jordan into global infrastructure and now accelerating agentic AI is the real story here. The medal is just the country catching up to what you've already done.
@bradmillscan Letting the agent drop files wherever it wants creates invisible tech debt that slows you down fast. Good structure keeps the project navigable, makes future changes cheaper, and actually helps the AI work better because it can find things without burning context.
@garrytan Opus 4.8 + OpenClaw is the first setup where the agent stops pretending and actually shows its work. Transparent fixes + real co-debugging is the difference between AI that slows you down and AI that multiplies output. Big for anyone shipping solo.
@beffjezos AI doesn't replace talent. It replaces people who treat it like a crutch instead of leverage. The gap between founders who actually get it and everyone else just became unbridgeable.
Most people use AI to answer questions.
Builders use AI to operate.
The difference is leverage.
If you’re building with agents (or want to), reply with what you’re working on.
We’re gathering the right people for the network.
What are you building? #8 @fdotinc@joypbuilds is building Stitch, a mini-vlog app for people who keep capturing life but never finish the edit.
This one hit me personally. I’ve been cutting videos almost every day recently, and honestly, editing is exhausting. You collect all these tiny moments, but then they sit in your camera roll or hard drive forever.
Stitch’s idea is simple: lower the pressure, reduce the choices, and help you turn daily clips into a finished mini-vlog.
“Less features is the feature” is such a good product philosophy.
Interview by @JoshTalksBiz
Camera by me
Thanks Joshua for helping make this one happen.
I’m also building https://t.co/TRBuTcr4U3, where your AI agent finds the right people and opportunities for you.