
A GAIA meetup on software quality, token cost and human effort. Daniel Nilebäck, co-founder of Voidleap, looks under the hood of coding agents: context management, agent scaffolding and choosing models for subtasks. Sponsored by System Verification and hosted by ai:gothenburg. Register on Meetup.
what it is
Getting an agent to write code is one thing. Getting software with the right quality, at a reasonable token cost and with a level of manual effort you can justify takes engineering.
Daniel Nilebäck explores how the harness, agent and tool definitions, and model choices shape your workflow. The technical talk offers a mental model for evaluating and improving your setup, regardless of which coding agents you use.
the questions on the table
- what is the difference between a harness, your instructions and your scaffolding?
- how do you balance software quality, token cost and human effort?
- why do context windows still matter, and why should compaction be a last resort?
- how can context offloading and different models for subtasks improve your workflow?
- how do you inspect what is in context and make scaffolding more token-effective?
on stage
- Daniel Nilebäck: co-founder of Voidleap, a Gothenburg-based startup developing Voidleap Code, an agentic IDE built on the company's own harness and agentic runtime.
good to know
- organized by GAIA, sponsored by System Verification and hosted by ai:gothenburg
- for developers and others interested in AI and agentic engineering
- beer and pizza at arrival, followed by the talk, Q&A and time for discussion
- bring questions about your own agent workflow
- register on Meetup
Who's talking
How the day runs
Where it happens
ai:gothenburg hub
ai:gothenburg hub, Masthamnsgatan 21, 6th floor
Heads up: ai:gothenburg doesn't have its own labeled button in the elevator yet. Just take the elevator to the 6th floor.
Made possible by
Save your spot.
Registration is handled by the organiser. Free, no payment required.
register on Meetup