Always Five Years Away
How 18 years of Hacker News predictions kept sliding the self-driving deadline
Key Insight
In 2010 the finish line was five years away. In 2015 it was still five years away. By 2025, the most common deadline had become the next five-year window.
HN did not lack optimism. It lacked a stable definition of "done".
What the Dataset Emphasizes
7,980 comments from 5,108 authors across 2007-2026. The language of deadlines and doubt repeats more than any single company.
Deadline Phrases
Company Gravity
Doubt Lexicon
How the Story Evolved
The Pioneers
- DARPA Grand Challenge halo and university prototypes
- Early Google experiments; maps seen as the missing piece
- Optimism framed as a regulation problem, not a tech problem
The Hype Machine
- Automaker timelines: 2014, 2015, 2020
- Tesla Autopilot launch and the rise of LIDAR roof racks
- The default promise: "2-5 years"
Reality Bites
- Uber fatal crash triggers a tone shift
- Waymo and Cruise move toward limited pilots
- HN asks: where are the cars?
The Long Road
- Geofencing becomes the unglamorous default
- Tesla FSD stays "beta" in the discourse
- Promise language shifts to "still" and "waiting"
The Optimists vs The Skeptics
Optimists (deadlines that slipped)
Volvo are planning to ship their first self driving cars in 2014.
What happened: 2014 arrived, and the deadline moved again.
GM's claim they would have a self driving car available for sale in 2015.
What happened: The five-year window survived another cycle.
the company's goal was to put a level 5 autonomous vehicle on the streets by 2020.
What happened: Level 5 stayed aspirational in later threads.
Skeptics (the long view)
We'll never see a Google self driving car commercially. It's a science experiment.
Why it aged well: The commercial rollout stayed narrower than early hype.
I am making a bold prediction that it will be decades before a self driving car can drive itself safely on Indian streets.
Why it aged well: Global edge cases kept the finish line far out.
Self driving cars are 15 years away.
Why it aged well: The 15-year horizon still felt unclaimed by 2026.
Reality Checks (2016-2020)
Yea, Google drove a blind man to Taco Bell in 2012 and still hasn't released a self driving car 4 years later.
I'll bet you money there's no legal self-driving car subscription like that in DC by 2023
Recurring Themes That Would Not Die
Level 4 vs Level 5
Level 4: 108 mentions (first 2014) | Level 5: 98 mentions (first 2016)
Level 4 autonomous vehicles are still years away. Level 1 - 3 cars can't legitimately be considered "self-driving" in any real sense.
LIDAR vs camera-only
LIDAR: 119 mentions (first 2007) | camera-only: 3 mentions (first 2024)
You can spot a Google self-driving Lexus by the big, spinning LIDAR on the roof. Teslas' somewhat-autonomous cars look like cars.
Edge cases never end
"edge case" appears 46 times (first 2015)
One can scale up the first 99% of an autonomous driving os pretty fast just using pattern recognition, but the devil is in the edge cases.
The trolley problem
Trolley problem appears once (2019) - more meme than metric
Use of high fidelity tech like LIDAR also causes another tendency, it lures engineers into thinking that it will somehow make tractable, a hard problem like the trolley problem.
Jobs and displacement
Driver/job language appears 869 times (first 2007)
I remember the transport minister of India claiming that he would ban self-driving cars to protect the jobs of drivers. No city in India is even close to ready for self driving vehicles.
2023+ Reflections: The Last 10%
Both technologies have come a long way and are quite impressive, but not quite there to actually be able to replace humans, and getting the last 10% of edge cases to work seems to require a monumental effort still.
Most of the criticism with self driving is around the over-hyped under-delivered "full self driving" from Musk -- it actually seems to be getting worse compared to 5 years ago.
Until I see a self driving car navigating through the streets of Bengaluru in the peak traffic periods, I would say it's still far far away.
The Verdict
Deadlines are stories, not schedules
HN kept using the same time horizon even as the industry shifted from demos to geofenced pilots. The calendar moved; the promise stayed.
Define "done" before forecasting
The debate kept oscillating between Level 4, Level 5, and consumer ownership. Without shared definitions, forecasts collapse into optimism.
Scale exposes the long tail
Every optimistic year was followed by edge-case reality. The last 10% of safety is not a rounding error; it is the product.
Generated by HN Zeitgeist - 20 years of tech discourse, analyzed. All citations verified against source data.