详细内容
Building the Global AI Workforce
Brainbase Labs is an applied AI research lab enabling AI in the workforce by building high-fidelity agent environments and tooling.
Our thesis is that as AI costs asymptotically converge to the cost of energy, the proliferation of AI employees will not come from best of batch vertical point solutions with SaaS margins, but rather from a single entity that will act as the workforce supplier of the entire global economy charging a small X% tax on each "worker minute".
Why is this possible?
Point solutions are not scalable. The unprecedented demand for forward deployed AI engineers is a strong sign towards enterprise AI adoption requiring a more flexible approach than the familiar one-size-fits-all SaaS model.
Vertical AI companies are a race to the bottom. The lack of differentiation and low switching costs for vertical AI solutions create a commoditized market where margins tend to zero in the long run.
Knowledge work is environment-constrained. We predict that the automation of a significant portion of economically viable knowledge work as it stands right now is primarily environment-constrained rather than requiring unforeseen leaps in intelligence.
Our Approach
We aim to build the global AI workforce by building generalist agents that live in high-fidelity virtual environments where they have access to the tools they need, followed by a global marketplace for hiring these AI workers salaried by the minute.
Reaching a true AI employee that is virtually indisinguishable from a remote human worker requires us to conquer three primary frontiers: planning, memory and always-on-behavior (AOB). We have already made significant advancements in planning which has allowed Kafka to achieve state-of-the-art performance on GAIA at a fraction of the cost without using reasoning models. We are currently hard at work on solving memory. If you'd like to help, please let us know.
Guiding principles
Bias towards human-like agents. Economically viable AI workers will have to work in settings designed for humans, which include using a keyboard and mouse, having an email and a phone number, and more.
Utility-like cost structure. AI workers will be charged per-minute with additional costs associated with bandwidth, surcharge pricing and more.
Bitter-pilled. High-fidelity virtual environments and capable agents in the workforce will unlock new S-curves in AI training by incentivizing reward labeling.
Products
Kafka, The First AI Employee
Stream
Play Video
0:00 / 3:54
Play
Currently unmuted. Volume set to 100 percent. Click to mute.
SettingsEnable Picture in PictureEnable Fullscreen
Kafka is the first AI employee.
Just like a real employee, Kafka comes with his own computer where he can code, run shell commands, read and write files and even use the browser to complete tasks. He also has his own email, phone number and Slack account so you can reach him anytime anywhere.
Kafka achieves state of the art performance on GAIA Level 3 at a fraction of the cost, demonstrating its ability to complete real world knowledge tasks.

We see Kafka as the first step in building a truly collaborative workforce where humans and AIs work together. We're already hard at work on the next generation of Kafka, and we can't wait to see how you put this one to work!
Join our waitlist today at https://usekafka.com and follow us @BrainbaseHQ on X for updates.
Hermes, Conversational Agent
Hermes is a special variant of Kafka which has been trained for low-latency high-volume conversational tasks.
Unlike Kafka, Hermes only has access to a special code runtime where he can run a Python-derivative called Based. Based is a language we have purpose built for conversational agentic flows and combines the reliability and speed of code with the naturalness of a generalist agent, allowing Hermes to handle thousands of concurrent conversations without hallucinating or losing track.
Hermes is currently used by everyone from Fortune 30 to airlines to handle thousands of interactions every day. If you'd like to try Hermes for your company, contact us.
Current Research Interests
Generalist Agents
- • Planning: Advanced decision-making capabilities
- • Memory: Long-term information retention and retrieval
- • Always-on: Continuous operation and availability
Core Research Areas
- • High-fidelity Environments: Agent environments for different tasks
- • Long-term Planning: Strategic decision-making over extended periods
- • Collective Memory: Shared knowledge across agent systems
- • Computer Use: Direct interaction with computer interfaces
Careers
We are actively hiring for technical positions. Join our team of researchers, engineers and AI workers working on enabling the global AI workforce.
We anticipate Brainbase to be one of the first organizations to have a hybrid human-AI workforce. Our goal is to have an exceptional team of no more than 55 humans and orders of magnitude more AI workers. As of July 2025, we are at 5/55 and are actively hiring for more.
Open Positions
Human Engineer
San Francisco, CA
Application Process
We are actively hiring for a variety of roles. To learn more and to apply to a role, send an email to hiring@usebrainbase.com.
Our application and pre-screening process is run entirely by a Kafka agent, who will answer your questions over email and will schedule a time to speak with you after evaluating your background and resume.
Kafka will then call you for a 15 minute interview, after which he will make a final recommendation to the team, at which point a human will contact you for the rest of the process.
To pick up a draggable item, press the space bar.
While dragging, use the arrow keys to move the item.
Press space again to drop the item in its new position, or press escape to cancel.
