In Search of the Legendary Ciudad Blanca#
Since the 1920s, several expeditions had searched for the legendary White City, or Ciudad Blanca. Spanish conquistador Hernán Cortés reported hearing “trustworthy” information about a region with “towns and villages” of extreme wealth in Honduras, but never located them. In 1927, aviator Charles Lindbergh reported seeing a “white city” while flying over eastern Honduras. The modern search for Ciudad Blanca is covered in the book The Lost City of the Monkey God by Douglas Preston which I read earlier this year (and highly recommend!).
In 2012, the ruins were first identified during an aerial survey of a remote valley in La Mosquitia, a vast region of jungle containing some of the last scientifically unexplored places on earth. This discovery was enabled by new technology: LiDAR (Light Detection and Ranging). LiDAR is a remote sensing method that uses pulsed laser light to measure distances and create precise 3D maps and models of the Earth’s surface, objects, and environments. The same technology now used in many electric vehicles today for self driving features, but fantastically more expensive for aerial surveys at the time in 2012.
Preston drew a sharp distinction regarding this enabling technology: LiDAR gives you discovery, not knowledge.
LiDAR reveals shapes, outlines, and patterns at superhuman speed. What it does not do is explain why a structure exists, how it was used, or what it meant to the people who built it. That work still requires humans on the ground, making judgments, forming hypotheses, and arguing with one another until meaning emerges.
GenAI is doing something eerily similar to software development today.
What GenAI is Undeniably Great At (and we under-credit it)#
Let’s be honest first. GenAI is not just “autocomplete on steroids.” It can:
- Generate working code faster than most humans can scaffold a project
- Explore vast solution spaces without fatigue or ego
- Refactor legacy code with mechanical patience
- Encode patterns across millions of repositories no human could ever read
I'm not joking and this isn't funny. We have been trying to build distributed agent orchestrators at Google since last year. There are various options, not everyone is aligned... I gave Claude Code a description of the problem, it generated what we built last year in an hour.
— Jaana Dogan ヤナ ドガン (@rakyll) January 2, 2026
In LiDAR terms, GenAI is phenomenal at revealing terrain. It shows us what is possible, what is common, and what is structurally sound. The mistake many teams make is assuming that visibility equals understanding.
It doesn’t.
What Human Developers Still Do that GenAI Cannot#
Here’s the provocative part: the most valuable parts of software engineering were never about typing code. Humans still uniquely:
- Decide what problem is worth solving in the first place
- Understand organizational context, incentives, and power dynamics
- Trade off speed vs safety, elegance vs pragmatism, local vs systemic impact
- Carry accountability when systems fail in the real world
- Create taste: knowing when code is technically correct but strategically wrong
LiDAR can tell you there is a structure in the jungle. It cannot tell you whether it was sacred, defensive, abandoned, or misinterpreted by history.
GenAI can write a microservice. It cannot tell you whether that microservice should exist at all.
The Dangerous Illusion: Productivity without Wisdom#
The real risk is not that GenAI replaces developers. The risk is that it floods organizations with discovery while starving them of knowledge.
We’re now in the era of slop code.
— Naval (@naval) January 5, 2026
When:
- Juniors ship code they don’t understand
- Seniors stop interrogating architectural decisions
- Velocity becomes a proxy for insight
You get systems that work until they suddenly, catastrophically, don’t. This is how technical debt becomes institutional debt.
The Uncomfortable Truth for Developers#
If your value is primarily:
- Syntax mastery
- Framework familiarity
- Boilerplate generation
GenAI is already better than you, and it will only get better.
If your value is:
- Systems thinking
- Judgment under uncertainty
- Translating messy human needs into durable technical decisions
Your importance is about to increase, not decrease.
The Real Parallel Preston was Pointing At#
LiDAR did not make archaeology obsolete. It forced archaeologists to become better thinkers.
GenAI will not end software development. It will end unexamined development.
The future belongs to developers who can walk the terrain GenAI reveals and turn discovery into understanding.
That was true in the jungle. It is even more true in code.
Discovery is cheap. Knowledge is earned.
Margin Notes#
The 2012 aerial survey of the Lost White City was followed by a 2015 expedition that explored one of the settlements discovered in the original survey. Douglas Preston covered this survey for National Geographic. This resulted in both the book and documentary The Lost City of the Monkey God:
Featured image credit: National Geographic