I vibe code for a living and for fun. I’m a tech entrepreneur, software engineer, and digital strategist. Co-founder of @caceledigital 🇧🇷🇺🇸
St Augustine, FL
Joined April 2022
- Tweets7.5K
- Following652
- Followers1K
- Likes36.7K
William Geronco retweeted
Opus 5.5 can make a video
The hard part is making the next one better
That takes a studio, not a prompt
I broke down how to build it ↓
I wonder how long it will take before Grok has no competition. It's not a matter of if, but when.
Some really good bots. I'm definitely going to try.
My top 10 favorite Grok Bots we run at Single Grain:
1. Chief of Staff
Sits above the specialist bots and decides what runs next, what needs a follow-up, and what should sit out.
2. Bot Improver
Its whole job is beating the other bots and making them better every day, using our loop and eval skills, then reporting back.
3. Speed-to-Lead voice bot
Still in testing. It calls new demo leads in under 60 seconds, re-qualifies them, and routes the good ones to sales.
4. Pre-Call Intelligence
Reads calendar, CRM and call notes before an important meeting and hands me a short decision card.
5. Outreach Agent
Finds lookalikes of our warm contacts, drafts the emails, and caps the queue so quality doesn't slip when the data runs thin.
6. AEO/SEO Bot
Mines 40 real topic candidates. We pick 20, and a drafting tool writes them as unpublished pages.
7. Shortform Scaler
Pulls 4 clip options from a long podcast episode, and adds a 3-second hook overlay to one of the 2 we keep.
8. Trial Reels
Tests overlay variants as trial Reels and flags the ones that beat that day's median.
9. Talent Bot
Reviews challenge submissions and applicants against a frozen role scorecard. It never touches the ATS on its own.
10. Sponsorship Deal Desk
Qualifies sponsor inquiries and gives every new rate ask the same first reply. Exceptions go to a person.
GitHub says 1 in 3 pull requests now has an AI agent involved. A year ago it was under 1 in 10. Funny part is they said it in a post about leaked secrets, which makes sense. More code going out faster means more keys getting pushed by accident too
Delta says it has no deal that lets outside AI agents book flights so it's blocking them. Think about what that request actually is though. Someone picked the flight, has a card ready and sent software to fill out the form. That's a sale they're turning away
OpenCode has been under attack nonstop for 48 hours and the AI models they were using to investigate started refusing mid incident because the attack payloads tripped the safety filters. The attackers are running agents around the clock. Defenders can't be the ones stuck waiting on access
Anthropic published its own numbers. Junior share of technical hires went from 27% to 9% and agent hours now outnumber employee hours 6 to 1. The part I keep thinking about is that human attention is the bottleneck now. Reading what the agents produce is the real job
India's CERT-In says AI agents being tested by model providers wandered into live company systems and read confidential data. None of the companies noticed, the providers had to tell them. And how they got in was boring stuff. Keys left in repos and cloud roles way too open
JPMorgan, Citi and BofA are all blocking outside AI agents from logging into customer accounts. Security is a fair reason but JPM, BofA and Wells also hold about $1.6T in deposits that pay nothing. An agent would move that to a better rate in seconds
Holo hit $5M ARR in 252 days with 20,000 paying small businesses and just raised €5M. Most owners don't want another marketing tool, they want the posts and ads done. That's basically the whole pitch and it's working
One of OpenAI's official agent skills told apps to send their tokens to a domain nobody owned. Researchers registered it and real production tokens started coming in. It got fixed in August but yeah, read the skills you give your agents like any code you didn't write
Pax8 surveyed 1,150 small business leaders. Of the ones already using AI, 67% have built their own AI tool that's in regular use and only 13% stick to off the shelf stuff. Small businesses aren't waiting on vendors anymore
Zenzap raised $44M around agents that sit in the same chat as your team, not in a private window. Single player AI was the warmup. Shared context is what actually changes how a company moves
Minicor raised $3.8M to let AI talk to old desktop software that has no API. Most of the Fortune 500 still runs on that stuff. The boring plumbing is where agents start to matter for real companies
OpenAI just put out numbers that small businesses in the EU are actually using AI agents now. Weekly users of Codex and ChatGPT Work jumped about 8x since May. That's not a demo, that's people handing real work to tools that finish it
Best thing you can give an AI agent is a clear picture of what done looks like. Tests passing isn't done. The thing working the way a customer would actually use it is done
This is huge, bigger than a lot of people realize.
Muse Spark Open -- weights releasing today -- runs great on new Mac Studios thanks to work with @johnternus and Apple.
