Just pitched Tieout at @join_ef (EF) Demo Day.
If you hate paying your auditor by the hour for their own inefficiency, come talk to us.
We are building the AI native audit firm. Starting with the 90,000 401(k) audits done every year. Then every audit.
If you are an investor who believes in AI native services, we are raising our seed round.
Email me at [email protected]
@buildwithsid for make design it your platform for me but if i would have searched for something like "make a graphic design" and for this you are not in first page
before we wrote history, we spoke it
for most of human civilization, it wasn't written down; it was spoken; generation to generation, for tens of thousands of years
now voice is becoming the interface to intelligence, and we’re building project mnemosyne for this chapter
visit https://t.co/p3tBfH6kzb to learn more
One recurring problem people are overlooking -- How do you share artifacts created by CLI agents?
The core problem is that CLI Agents are local first!!
Making artifacts available online, shareable, and consumable by both humans and other agents is surprisingly hard. It's fundamentally an Agent Experience problem.
There are workarounds today, but none are designed for CLI Agents. You end up dealing with attachments, emails, Google Drive links, permissions, and authentication.
We built our first product to solve this by making pdfs, pngs/jpegs, mp3s, mp4s, xlsx/csvs, mds, and ppts created by your CLI Agents instantly available online.
📁 Codex Report Gen Agent
└ 📄 latest-report.pdf 🟢 Online
📁 Hygen Hyperframes Agent
├ 🎥 final-video.mp4 🟢 Online
└ 🖼️ frame-1.png 🟢 Online
📁 Gemini Agent
├ 🖼️ logo.png 🟢 Online
└ 🖼️ face-swap.jpeg 🟢 Online
@Supafork Artifacts makes artifacts instantly available online with a shareable URL that both humans and agents can access. Not attachments. Not emails. Not Drive links. Just a simple, agent-native way to share artifacts through a language Agents understand.
Excited to make it available to our beta users tomorrow. Join the waitlist on https://t.co/UgNoD6Iqoi
Most people build a swipe file by saving whatever has a lot of views.
That's the wrong filter.
The number that matters isn't total views. It's how far a video overperforms relative to that specific channel.
outlier score = video views ÷ channel's median views
A channel that normally pulls 15K views posts something that hits 750K. That's a 50x outlier. The channel found a nerve, not just a good day.
Here's the threshold I use: → Under 10x — ignore it, normal variance → 10x-30x — worth studying, something's working → 30x+ — swipe-worthy, this is a format to model this week
The trap is chasing raw numbers. A 3M-view video on a channel that averages 6M didn't teach you anything — that's just Tuesday for them. But a 750K video on a 15K-average channel just handed you a map.
Stop collecting big numbers. Start collecting big deviations.
@nestymee@frederickjames are these videos doing good now?
what are your learnings for older not working and why these ai variations are doing good(if doing good)