Newsletter: Let AI Hear It
AI Agents, Bubble Talk, and Finally: it's ok to yell at ChatGPT
Brixo Latest
We just published an AI Agent Directory, which features a curated database of AI Agent companies. The goal is to provide a resource for those who want to learn about agents, and find an agent solution.
We sourced this list by reviewing research and market analyses focused on agents. The goal was to highlight companies building real AI agents, not generic AI tools.
We reviewed public datasets, reports, and venture lists, then evaluated each company individually to confirm it develops agent-driven products that plan, reason, and execute autonomously. Some well-known AI vendors are not included for this reason.
After narrowing the list, we compared firmographic data, funding, and product updates to ensure the information was current.
We’ll continue to publish lists showcasing each category over the next few weeks.
What stood out
What we read/listened/watched
Is there an AI bubble?
Gavin Baker, managing partner at Atreides Management, shared a unique perspective on the AI Bubble talk. Spoiler: NOT in a bubble
From idea to $650M Exit: Lessons in Building AI Startups
A great talk by Jake Heller, co-founder of Casetext, which sold to Thomson Reuters for $650M. Casetext is a legal AI company.
Ruslan Belkin on how Agents Can Fix Themselves
Ruslan, head of engineering at Reid Hoffman’s inflection.io, shares a unique perspective on how an intelligence data flywheel as the best technique for improving agent behavior — we agree.
Vercel CEO Guillermo Rauch on the future of Agents
Fast-forward to 2:26 to hear Vercel CEO Guillermo Rauch talk about the Agent Cloud and his perspective on the direction that we’re going in.
Final brick
It’s ok to yell at ChatGPT!
Penn State researchers discovered that being rude to ChatGPT makes it 5% more accurate.
They tested 250 questions across tones from “Very Polite” to “Very Rude.”
Results jumped from 4.8% as politeness dropped.
Apparently kindness is not the key. Please and thank you confuse the model.
The more direct and rude, the better it performs.
We all can blow off some guilt-free steam at ChatGPT in the name of optimization!
