The rulebook outgrew the humans

Civil engineers did not get slower or worse. The corpus of codes, standards and required documentation grew past what any one person can hold in working memory. That is a system problem, and it has a system solution.

Two people descending a concrete staircase in a daylit atrium

The engineers did not get slower

No. The one long-running measurement of how engineering firms actually spend their hours, Deltek's annual Clarity study of architecture and engineering firms, shows billable utilization at the top of its ten-year trend. Engineers work as much, and as billably, as at any point in the series. Whatever is slowing projects down, it is not the people.

What changed is the material. Every failure adds a rule, every rule requires documentation, and rules rarely retire. So document volume went up, decade after decade, while human working memory stayed where biology left it. Nobody decided that the sum of individually reasonable requirements should exceed what one person can hold. It happened one justified paragraph at a time.

The rulebook grew, measurably, at every level

Measurably, and on the record. The Eurocodes, Europe's common design standards for structures, went from 10 standards in about 58 parts in their first generation to 11 standards in about 65 parts, plus a dozen supporting technical specifications, in the second, on the EU Joint Research Centre's own count. The growth is not padding: the second generation adds entire new codes, one for structural glass and one for assessing and retrofitting existing structures.

Sweden's AMA Anläggning, the reference work behind most civil-works specifications, grew from 742 pages in the 2007 edition to 926 pages in 2023: a quarter thicker in sixteen years. The same drift runs through the American codes. Ron Klemencic, chair and CEO of the structural firm Magnusson Klemencic Associates, warned in ASCE's Civil Engineering magazine that engineers risk becoming "compliance officers":

The scope of such codes has expanded from tens of pages of provisions to hundreds of pages.

Ron Klemencic, Civil Engineering (ASCE), September/October 2025.
Measure Change Source
Eurocodes, first to second generation 10 standards, 58 parts → 11 standards + 2 TS, 74 parts EU Joint Research Centre
AMA Anläggning, 2007 → 2023 742 → 926 pages Svensk Byggtjänst
Swedish companies' regulatory cost, 2024 SEK 389 billion per year Tillväxtverket
UK infrastructure consent decision, 2012 → 2021 2.6 → 4.2 years House of Commons Library
US code changes since 2012, share of new-apartment cost 11 percent NAHB survey, via Cato

The accumulation costs money, and mostly time

Sweden's enterprise agency Tillväxtverket puts companies' total annual regulatory cost at 389 billion kronor for 2024, six percent of GDP, and the figure has risen year on year. That covers all sectors, not construction alone, but it sets the scale: compliance is one of the economy's largest line items, and construction rules such as climate declarations appear among the named cost drivers.

In infrastructure, the cost arrives mostly as time. In the United Kingdom, the average consent decision for a nationally significant infrastructure project took 2.6 years in 2012 and 4.2 years in 2021, 62 percent longer within a decade, according to the House of Commons Library. In the United States, a trade survey cited by the Cato Institute found that building-code changes adopted just since 2012 account for 11 percent of the cost of building a new apartment.

None of this is a law of nature. US federal environmental reviews recently got faster, after Congress set a two-year deadline in 2023. Timelines respond to how the system is designed. That is worth keeping in view, because it means the accumulation is a design outcome, not a fate.

A system problem has a system solution

Now the uncomfortable part: no individual regulation is wrong. The glass code exists because glass fails differently from steel. Climate declarations exist because concrete carries carbon. The consent process exists because infrastructure runs through other people's land and water. Pick almost any single rule and it can name the failure it answers. The problem was never one rule. It is the sum, and the sum passed the limit of human working memory some time ago.

That rules out the two standard fixes. "Cut red tape" stalls, because every rule defends itself when examined one at a time. "Hire more engineers" does not scale, because each new engineer brings the same working memory as the last, and the coordination between them becomes its own document stream. A corpus problem needs a reader that scales with the corpus. Machines can now hold an entire rulebook and its cross-references at once, and check work against all of it, every time, without tiring. The engineers never outgrew the work. The rulebook outgrew the engineers. What is new is that this has become a solvable condition.

  1. Deltek, Clarity Architecture & Engineering Industry Study, 2025.
  2. European Commission, Joint Research Centre, "Evolution of the EN Eurocodes"; part counts from JRC report JRC144386 (2025).
  3. Svensk Byggtjänst, AMA Anläggning 07 and AMA Anläggning 23, page counts per the publisher's catalogue data.
  4. "Taking a look back at Vision 2025 and exploring what's ahead", Civil Engineering (ASCE), September/October 2025.
  5. Tillväxtverket, Regler som påverkar företagens kostnader och konkurrenskraft 2024, 2025.
  6. House of Commons Library, Planning for Nationally Significant Infrastructure Projects, briefing SN06881.
  7. Emily Hamilton, "Reforming US Building Codes", Regulation, Cato Institute, Winter 2024–2025.
Benjamin Glaser Co-founder at Yesper. Writes about AI and the industry that builds the world. benjamin@yesper.ai

Yesper is the AI civil engineer for construction and infrastructure. AFRY, COWI, NRC Group and other Nordic firms use it to halve the time on a study, rerun calculations in minutes, and catch errors that would otherwise slip through. Get in touch if you'd like to see what it can do for you.

Book demo