EDL employee Alessandro built a benchmark to test whether machine-readable documentation fixes agent tokens going to re-reading a codebase.
Public source
Publisher name
Public post
94% of an agent's tokens can go to re-reading a codebase before it writes a single line. Alessandro built a benchmark to test whether machine-readable documentation actu…
Company
EDL
Copenhagen's top B2B design system specialists
- Industry
- Software Development
- Location
- Copenhagen, DK
- Company size
- 11–50 employees
About EDL
EDL is a strategic design engineering studio. We design and launch AI-ready design systems for B2B SaaS teams that compete on product quality. We work with product-led companies where design is not a surface layer, but a core part of how the business performs. That usually means complex B2B SaaS, growing teams, and real constraints around speed, scale, and consistency. Our work sits at the intersection of UX research, product design, design systems, and design engineering. We embed with product and engineering teams to turn insight into systems, and systems into measurable outcomes. No handoffs. No theatre. No opinions without proof. We help teams move faster without breaking quality, align design and engineering, and build products that users actually want to use. Design isn’t decoration. It’s infrastructure.
See moreLatest activity
Latest activity from EDL
3 signals
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
enola
enola indexed 91 public repositories three times, including one cold run and two warm runs, to benchmark its capabilities and limits.
Research & Knowledge
E1
E1 tested eight tools against 67 real production bugs this July, achieving a best F1 score of 47%
Research & Knowledge
NEC Laboratories America, Inc.
NEC Laboratories America, Inc. published a paper titled Escaping Whack-a-Mole on July 11, 2026, optimizing documentation as repo-specific playbooks for coding agents.
Research & Knowledge
Storyblok
Storyblok ran an evaluation where installed agent skills scored exactly the same as no documentation at all, with both 53%
Research & Knowledge
This Dot Labs