2026 Predictions Scorecard | Hacker News Reader

2026 Predictions Scorecard

rodneybrooks.com

64 pointscalvinfo39 comments

Of the predictions I read, I found that the author engages in pretty heavy handed rules lawyering in order to make their predictions accurate.

For example, the author takes the stance that current self driving cars (Waymo, Zoox) do not count as self driving. The justification being that a human operator is involved some small fraction of the time.

By law, Waymo must report disengagements in California. In 2024, Waymo had ~10 thousand miles driven per disengagement, Zoox had ~28 thousand miles driven per disengagement [1]. I would say that this rate of human intervention qualifies as self driving.

[1] https://thelastdriverlicenseholder.com/2025/02/03/2024-disen...

On the contrary, it’s the companies doing the lawyering. A disengagement is when the vehicle reverts fully back to manual control. Tele-operation does not count as a disengagement, and the frequency of tele-operation intervention is a closely guarded industry secret.

You seem to have some deeper insight into this - in your estimation, how often does tele-operation (even a small correction) take place?

There’s been reporting on this in several mainstream publications that was accurate as far as the systems I worked with. Unfortunately I don’t want to dox myself on here, so unsatisfyingly the best I can offer is “trust me bro”.

The tele-operation is also kinda vague because as I understand it, with Waymo at least, they are not turning a steering wheel and pushing pedals at HQ, they are saying "Pull over here" etc.

Oh interesting, I had figured tele-operation would count as a disengagements.

Looking into reports you mentioned in a child comment, CNN reports Cruise needed human assistance every ~5 miles [1]. And I certainly wouldn't call a system that needs assistance every ~5-10 minutes Level 4 self driving.

Subjectively, it appeared Waymo was significantly better than Cruise in 2023 but without data it's hard know what that means in terms of human intervention.

If Waymo needed human assistance every 10-20 minutes, I would agree that it also doesn't qualify as Level 4 autonomous.

[1] https://www.cnbc.com/2023/11/06/cruise-confirms-robotaxis-re...

Predicting well absent lawyering is really hard! If someone else wants to try I warmly recommend starting with e.g. the ACX 2026 prediction contest: https://www.metaculus.com/tournament/ACX2026/

> Predicting well absent lawyering is really hard!

The author engages in rules lawyering of the evaluation of the predictions. The original predictions are clear.

Another example of this is the author's prediction that no robot will be able to navigate around the clutter in a US home, "What is easy for humans is still very, very hard for robots."

The author evaluated this prediction as not being met, "...I don't count as home robots small four legged robots that flail their legs quickly to beat gravity, and are therefore unsafe to be around children, and that can't do anything at all with their form factor besides scramble".

The author added constraints not in the original prediction (safe around children, must include a form factor able to preform an action, ...) then evaluated the prediction as accurate because no home robot met the original constraint + the new constraints.

Those robots are not "navigating AROUND the clutter", there are no consumer-deployed object recognition physics models that let a robot say "that's a ball that will roll, that's a shirt that will tangle, that's a sheet of paper that may slip", they're just charging chaotically through it. If you allow skittering robots, do you exclude a 90s RC car with the trigger taped down?

I would agree with you if the author wrote, "I don't count as home robots small four legged robots as they don't navigate around clutter, they just go over it" but the author didn't write that.

The author quoted two constraints (safe around children, must include a form factor able to preform an action) not specified.

The author projected that a lab demo of capabilities would not occur. I don't see safety for children as necessary for a lab demo.

I read the prediction and I see it meaning "I don't think we'll have tech for useful house robots in the average American Family Home"

The natural interpretation (for me!) was predicated on navigation - implying consideration and appropriate response to the clutter. Not merely ignoring it by being robust to the problems it engenders to movement/balance/etc.

As somebody who has been reading them since the first year, I think you have it wrong. That self-driving prediction was always about Level 5 autonomy. What's changed between now and then is that we've basically stopped talking about that, instead accepting intervention-as-a-service companies as self driving.

Well, ans we're talking about within very specific locales.

Honestly, Brooks--who has been presented and self-presented as something of a skeptic with respect to autonomous self-driving--looks like something of an optimist at this point. (In the sense that your kid won't need to learn to drive.)

The author quotes that their predictions are for Level 4 autonomy:

> The definition, or common understanding, of what self driving cars really means has changed since my post on predictions eight years ago. At that time self driving cars meant that the cars would drive themselves to wherever they were told to go with no further human control inputs. It was implicit that it meant level 4 driving. Note that there is also a higher level of autonomy, level 5, that is defined.

Thanks for the correction!

> By late 2024 the biggest numbers that had been factored by an actual digital quantum computer had 35 bits (citing https://arxiv.org/pdf/2410.14397v1 )

This is incorrect. The cited reference says "N <= 35". That N is the number being factored, not the number of bits in the number. Also, footnote a of that paper points out (correctly) that the circuits that were used likely needed knowledge of the factors to create (e.g. as explained in https://arxiv.org/abs/1301.7007 ). As far as I know, only N=15 has been factored on a quantum computer in a no-shenanigans way.

It's conceivable that current ion trap machines could do a no-shenanigans N=21.... but anyone judging progress in quantum computing by largest-number-factored is looking at the wrong metric (for now). You won't see that metric move meaningfully until quantum error correction is done spinning up.

Related. Others?

Predictions Scorecard, 2025 January 01 - https://news.ycombinator.com/item?id=42651275 - Jan 2025 (185 comments)

Rodney Brooks Predictions Scorecard - https://news.ycombinator.com/item?id=34477124 - Jan 2023 (41 comments)

Predictions Scorecard, 2021 January 01 - https://news.ycombinator.com/item?id=25706436 - Jan 2021 (12 comments)

Predictions Scorecard - https://news.ycombinator.com/item?id=18889719 - Jan 2019 (4 comments)

My Dated Predictions - https://news.ycombinator.com/item?id=16078431 - Jan 2018 (50 comments)

Reading the comments here about lawyering to make the prediction seems accurate, I have to say for me the value of prediction is not firstly about its binary accuracy, but more about the insights I get from the prediction and adjustment process. As our language is always limited, putting everything in a binary yes/no will almost always result in some dissatisfaction. We learn never much from binary values but from the gray values and how we must change our judgements to adapt. Perhaps that’s why punishment can cause a reaction but not deep learning and good educators always strive for insights and self-correction.

Useful predictions should also not in black or white but should be presented with an uncertainty, percentage of confidence if one can. It helps one to adjust ones prediction and confidence when new facts come along and I argue every serious predictors should do that.

"2. Self Driving Cars. In the US the players that will determine whether self driving cars are successful or abandoned are #1 Waymo (Google) and #2 Zoox (Amazon). No one else matters. The key metric will be human intervention rate as that will determine profitability." - I love that he's not mentioning the speculation company of the century. We don't have to mention it either.

Tesla? I don't love the company or the owner, but it seems silly to completely dismiss them so early on, relatively speaking. Self driving has been a decades long effort; even though I am heavily in favor of Waymo, some speculation towards Tesla's path seems fair. At the same time, I agree with the article here:

> Tesla (owned by Tesla) has put on a facade of being operational, but it is not operational in the sense of the other two services, and faces regulatory headwinds that both Waymo and Zoox have long been able to satisfy. They are not on a path to becoming a real service.

Given Tesla's abysmal track record on keeping their promises I feel like it is justified to dismiss them, at the risk of being surprised if they do make it.

Both of these can be true at the same time:

* Elon has been making wildly exaggerated and over-optimistic claims for a decade and continues to do so

* Tesla has recently made huge strides in capability and has a clear path to full autonomy

And to be fair, many other car companies also promised self driving cars, e.g. Audi in 2014 promising driverless cars by 2016 [1]. It's just that Tesla is still executing on the promise whereas many other carmakers have fizzled out on their ambitions. As the Rodney Brooks article itself mentions,

> As a reminder of how strong the hype was and the certainty of promises that it was just around the corner here is a snapshot of a whole bunch of predictions by major executives from 2017.

[1] https://www.digitalspy.com/tech/a610930/audi-promises-to-del...

Given that fully driverless Model Ys and Cybercabs have been spotted going around Austin, I find that the "they are not on a path to becoming a real service" is a little too strongly worded.

Tesla's commitment to no lidar means they aren't even in the race.

Seems overly reductive, both supply and demand will determine what happens. So far demand for Waymos seem fine, they can stimulate it way further by lowering the prices. The problem is on the supply side, specifically unit price economics. Intervention per mile is just one part that goes into profitability and I doubt it's biggest one. I would estimate the costs to be in this order - vehicle cost, maintenance (and vehicle longevity), human intervention, charging, fleet management (cleaning, etc), and regulatory environment.

In particular, Jaguar Waymos are over 150k a pop. It seems far fetched that any of them will make ever break even. New generation is reportedly $75k per vehicle which is significantly better. I could not find any data for Zoox vehicle cost, but given how few of them there are it's a non-player.

Finally the elephant in the room. Outside of camera vs lidar holy war, Tesla seems well positioned to dominate supply side of the equation if the demand shows up. Robotaxis are reportedly under $35k, they own the factories and know how to build more, they also own the maintenance side.

You can build a GMC panel van that seats 12 for about $20k, I don't think vehicle cost is a significant hurdle.

You can’t build a self driving GMC panel van with non-Tesla tech for $20k.

(Or, probably, with Tesla tech. But you definitely can’t do it without.)

> I love that he's not mentioning the speculation company of the century. We don't have to mention it either.

The word 'Tesla' appears 17 times in the article.

What’s even crazier is those fringe ideas from back in the day now have the compute power to actually execute as designed and really interesting things emerged.

I feel like we are on the verge of it, some say we already had it, but the singularity where hardware, learning, software, sensors and vision, all collide with enough headroom to make real-time our time. Now if we could only solve for our monkey brains.

So the shor's on real qc is still only 35.

And the record is N=35=7x5, that's 6 bits not 35 bits as the author is saying... Maybe he'd revise his prediction on QC if he knew?

35bit was on simulated quantum computer (on a classical supercomputer)

Prediction markets are superior for this stuff since one isn't self-evaluating and most prediction outcome evaluation is dominated by self-identity.

Flying cars are nearly ready. That prediction of it not happening until 2036 is going to be set false the soonest. Alef Aero cars just started manufacturing.

https://www.flyingmag.com/california-firm-first-flying-car-p...

> According to Dukhovny, a Model A prototype has been flying for years. However, there are no videos of the vehicle making the transition from a car to an aircraft. It will need to prove that capability, as well as its adherence to the rules of both the road and the sky.

> Dukhovny said the Model A will initially be certified as a “low-speed vehicle” on the ground, limiting it to about 25 mph on public roadways.

Sounds like they're just barely doing enough to meet the definition of "car" there, if it works at all. You'd think a video would at least exist if these are in production.

This sounds more like a personal aircraft where you can fold the wings in and propel it with a motor, not what people think of as "flying cars"

In context with the requirement of

"A flying car can be purchased by any US resident if they have enough money."

and the prediction level being

"There is a real possibility that this will not happen at all by 2050."

This production fulfills his requirements of the prediction becoming true much sooner than expected. He could go get one today with unlimited money.

> started manufacturing

There is a large graveyard of EV companies that hit this milestone. And they had a relatively easy problem space as well as a massive addressable market.

>Yes, I did say that something new and big that everyone accepted as the new and big thing in AI would come along no earlier than 2023 [...]

Lol, what a joke. I'd rather get someone to read my hand and tell me what's going to happen.

- Vulcano from Canary Island erupting

- Another one from Italy too

- Weird space event

- Odd geometry based discovery on Cosmologics

- Also, DNA's codons' layout will have something to say too

- Rust on Linux takes over Intel iGPU drivers (sadly). Tons of Linux distros either get OpenBSD's Xenocara or NetBSD's base X.org to get stuff working.

- Trump is impeached and declared mentally challenged due to an age related dementia in order to avoid prison time (and to avoid a global economy crash).

Reply on news.ycombinator.com