Developers are now “whispering to their agents” to not distract colleagues.
@steph_palazzolo reports: “I talked to a startup called Basis where basically if you walk into their office, there are like dozens of people just sitting at their desk with these sorts of mic’s that you might see at a podium.”
“And they're just whispering and talking very quietly into the microphones.”
Listen to our cofounder @mitch_troy speak about the applied ml research we're doing toward closing the loop with self-improving agents
Full episode linked 👇
AI Evolution: Self-Improving Agents & Future Possibilities
NEW EPISODE TOMORROW!
New episodes every week. Find us wherever you listen, and catch the video version on YouTube and Spotify.
https://t.co/TTB20CsPPv @hedgineer
https://t.co/8ceOnDBe2i
@detroitcoder
The term harness is too bloated. Everything is a “harness”. I propose a new concept: operating environment.
A harness is the execution layer. The code that orchestrates the model, memory etc. so Claude code is a “harness”
An operating environment is the totality of the world the agent lives in. This includes the harness BUT also includes all of the context the model learns from in the course of a trajectory. (Or as I like to think of it “runtime training data”)
So Claude Code or Codex would be a harness, but the Basis codebase when Codex runs on it is an operating environment.
Arguably this distinction is even more important for applied ML companies like Basis that own the operating environment for our agents and need to be very thoughtful on that context and meta behavior engineering
Excited to announce our collaboration with @BakerTillyUS
Baker Tilly is rolling out Basis across accounting and finance delivery within its managed services practice, helping its teams spend more time on the analysis, advice, and client work that matter most.
We’re proud to work with the Baker Tilly team as they continue strengthening a client experience built around consistency, visibility, and faster access to decision-ready insights.
Read more: https://t.co/YfcenB7kYS
GPT-5.4 is our most factual and efficient model: fewer tokens, faster speed.
In ChatGPT, GPT-5.4 Thinking has improved deep web research, better context retention when it thinks for longer—and oh—you can now interrupt the model and add instructions or adjust its direction mid-response.
Steering is available this week on Android and web. iOS coming soon.
We've raised $100m at a $1.15b valuation from @Accel, @GVteam, and existing investors to accelerate deployment of the most capable and accurate accounting agents across CAS, tax, audit, and advisory.
Basis is used by 30% of the Top 25 accounting firms and dozens more across the Top 150.
Today we're announcing the first accounting agent to complete a business tax workbook end-to-end.
Our focus on production-grade, long-horizon agents means that 12 months from now, the work Basis handles will make even this look routine.
We're looking for a few very intense people who want to build at the frontier.
At Basis, we built an agent that autonomously resolves more than 78% of reported bugs on the first pass.
It's such a good investigator, we called it Clueso.
@trybasis is an incredible place to work, and these kind of messages are really rewarding for me personally. Our motto has always been:
"We don't care about the perception of winning, we care about actually winning."
Buuuut perceptions matter. So we're hiring a a Narratives & Comms Lead! JD below.
For Halloween, the team secretly dressed up as our CEO Matthew Harpe with a pink polo and khakis.
Congratulations to the look-alike contest winner @matthew_busel