Tokyo Tech Talks #5: Human in the Loop
Join us on June 23, 2026 at Build+ in Ebisu for an evening of talks and conversation about what the human is for — taste, judgment, and oversight in a world of software that acts on its own.
We’re excited to announce Tokyo Tech Talks #5.
Last time, we closed the loop and let systems act, observe, and adapt on their own. This time, we put ourselves back inside that loop and ask a harder question: when software can act without us, what is the human actually for? The theme for this event is Human in the Loop.
Event Details
- Date: Tuesday, June 23, 2026
- Time: 7:00 PM - 9:00 PM JST
- Location: Build+, 6F, MARIX Ebisu Bldg., 4-4-6 Ebisu, Shibuya, Tokyo
- View on Google Maps
Schedule
- 7:00 PM - 8:00 PM: Talks — Short presentations from builders working alongside systems that act on their own
- 8:00 PM - 9:00 PM: Drinks & Networking
The Theme: Human in the Loop
The phrase “human in the loop” usually shows up as a safety valve — a person who approves, corrects, or steps in when the machine is unsure. But as software starts to act on its own, the human’s role is changing from operator to something harder to name. We set the goals. We judge the output. We decide what is good. We carry the taste the system doesn’t have.
That role is easy to lose. The same automation that frees us can quietly push us out of the loop entirely — until the only thing left to do is rubber-stamp a decision we no longer understand. The interesting design problem is not how to remove the human. It is how to keep the human meaningfully in the loop: present where judgment matters, absent where it doesn’t.
What We’ll Explore
This event brings together people building systems where the hardest question isn’t what the software can do, but where the person belongs.
Oversight Without Rubber-Stamping
When an agent proposes ten actions a minute, “review each one” stops being real oversight. We’ll look at how builders are designing approval, intervention, and escalation so that human attention lands where it actually counts.
Taste as the Bottleneck
Models can generate endlessly. The scarce resource is judgment — knowing which output is good, which is off, and why. We’ll hear from people whose work has shifted from producing things to steering and selecting them.
Working Alongside Software That Acts
Pair programming with an agent, designing with a model, debugging a system that rewrote itself overnight. We’ll look at what genuine collaboration with autonomous software feels like, and what it asks of us.
Who Should Attend
- Engineers and developers building or working alongside agents and automated systems
- ML and research engineers thinking about oversight, evals, and control
- Product and design people shaping how humans and software share the work
- Anyone curious about what stays human when software can act on its own
Join Us
If you’re interested in attending, register here.
Come spend an evening with people thinking seriously about where the person belongs in systems that increasingly run themselves. The talks will be short, the demos concrete, and the conversation open-ended.