Free with a sign-in.
The best 11 min of 22
Eyes-Off Autonomous Driving Arrives by 2028 | Brandon Basso
Brandon Basso argues that end-to-end driving AI can learn safety directly, while diverse data and validation remain essential to autonomy.
MTS · Sophia Dew
- Brandon Basso is head of simulation and validation at Wayve.
The best minutes
- 12:42Wayve learns safety instead of relying on checkersPro pick 3 min
- 8:13Wayve combines road data, OEMs, and simulationPro pick 5 min
- 15:48Wayve expects a shift toward eyes-off driving by 20283 min
Every moment plays free. PRO marks a write-up that's in Pro.
Also worth watching
Skip the opening company overview, the Lovable ad, and the closing sponsor reads.
Full analysis · 5 min read
Real Alpha · Nov 8
Brandon Basso says eyes-off shift may come in 2027 into 2028
- Basso expects a major shift toward eyes-off L3 driving in 2027 into 2028.
- His case rests on Wayve's safety, data and validation approach, but the episode reports no measured safety results.
- Teams should treat the window as a thesis to test, not a product commitment.
Free with an account
Read the rest of the analysis
6 sections, the 3 other voices it tests Basso against, and what it means for you. Every episode we cut gets one.
Check your inbox
We sent a link to your address. Click it to finish signing up. Not there in a minute? Look in spam.
or
Signed up before? The same buttons sign you in. The episode, its moments and the transcript stay free.
Basso expects an eyes-off shift by 2028
Basso forecasts a major shift in 2027 into 2028, from hands-off L2+ assistance toward eyes-off L3. He describes L3 as a system where occupants can do other tasks while the system carries safety responsibility.
That is an industry-level expectation, not a dated Wayve launch announcement. Wayve works from L2/L2+ assistance toward L4 robotaxis, and it has a supervised public pilot in London.
Brandon Basso leads simulation and validation at Wayve. He is presenting his company's technical case, not an independent audit.
Kai Yu, Horizon Robotics' founder and CEO, also mapped hands-off, eyes-off and mind-off driving. He did so on In Good Company in September 2026.
Yu's episode discusses the compute, data and models needed. That is a second industry roadmap, not evidence that Basso's forecast is proven.
“move beyond L2 plus to L3 systems”
Our read
Treat the timing as a forecast, not an availability date. The distinction matters when planning vehicle programs, budgets or customer promises.
Eyes-off systems require stricter safety proof
Basso says eyes-off systems require much stricter safety guarantees than hands-off assistance. He does not tie that requirement to a specific validation date.
He separates building a driver that can outperform humans from proving that it works. That split matters because a capable demonstration is not proof across weather, traffic rules and unusual events.
Dmitri Dolgov, Waymo co-CEO, discussed this on Minus One in October 2026. Open-loop tests miss how actions change the world, so his account calls for closed-loop validation.
That comparison makes the missing detail concrete. The episode does not say which closed-loop tests Wayve uses or what results they produce.
It also reports no test mileage, comparative safety rates or pass thresholds. That absence does not show Wayve lacks such evidence, but this interview cannot establish it.
Our read
The forecast rests on a proof burden the episode explains but does not show Wayve has cleared. Evidence must cover changing situations, not just a successful drive.
Four driving dimensions resist one score
Basso says no single metric grades driving quality. Wayve uses many metrics and tries to align them with human expectations.
For instance, he says traffic signs and rules in Japan and Germany differ. That makes compliance specific to local context, alongside safety, comfort and competence.
For example, Basso says a lane change should happen in a way people expect. That competence test sits alongside safety, comfort and compliance, not inside collision avoidance alone.
The four-part framework gives product and validation teams a wider checklist. But the interview provides no metric definitions, weights or thresholds for a release decision.
“we try our very best to make sure that those metrics are aligned with human expectation”
Our read
The framework prevents one score from standing in for driving quality. Its practical value depends on definitions and release gates Wayve does not share here.
Local variation widens the data burden
Construction scenes show how ordinary-looking situations become geographically specific. Basso presents examples from Japan and the UK involving signs, different characters and uncertain worker signals.
Drivers must parse language and infer what signs mean for a safe next action. That is a stronger demand than collecting large volumes of familiar road footage.
Byron Boots made the case on The Shawn Ryan Show in September 2026. He said data should span environments where a system may eventually drive.
That supports coverage as a design requirement, though Boots also notes new surfaces may need fresh data. Wayve's own collection, OEM partners and licensed sources broaden its inputs.
Basso says expert drivers provide human reference data too. Still, the interview gives no coverage map, rarity target or test for unfamiliar local conventions.
Our read
Coverage is a product requirement, not just a training input. The open question is whether the data mix captures rare local situations before deployment.
Simulation may scale data, but its share is unknown
Basso first says simulation should generate data primarily over time, then qualifies it with maybe not primarily, but over time. That makes simulation a direction, not a quantified supply plan.
Wayve also collects data itself, works with OEM partners and licenses data from vendors. Basso says partners have richer data, while internal operations target geographic and vehicle diversity.
For example, he says licensed dashcam footage might let Wayve add a highway event from ten years ago. He frames that as a possibility, not a completed addition or demonstrated gain.
Neither source counts nor simulation's added coverage are reported. Simulation volume alone would not show useful safety coverage; evaluation must show what it adds.
Our read
Simulation may broaden the data mix, but the episode offers no basis for budgeting around a particular contribution. Measure added coverage and effect, not volume alone.
Wayve learns safety alongside lower-level safeguards
Basso says Wayve treats safety as intrinsic to its main driving model, rather than an external checker. He contrasts that design with modular or hybrid stacks that rely on separate subsystems and checks.
Map-based checks can become brittle when a sign is temporary, moved or attached to a bus. Basso says checker systems can ultimately limit performance, but the episode offers no comparative test.
Wayve's demo represents future safe and unsafe driving areas, then uses that prediction to guide trajectories. But Basso also lists automatic emergency braking and lane-keeping systems among lower-level protections.
So the distinction is not AI instead of safeguards. It is learned safety in the principal driver alongside familiar safety systems.
That architecture is a design choice, not proof of better outcomes by itself. Without comparable results, buyers cannot tell whether learned safety generalizes better or creates new failures.
Our read
The episode explains Wayve's architectural preference, not a measured advantage over decomposed systems. A release decision still needs comparative evidence across the situations the system must handle.
What it means for you
- Define release gates. Write separate acceptance tests for safety, comfort, compliance and competence. Record thresholds and tradeoffs that would block release.
- Measure coverage gaps. Track which locations, vehicle types and unusual scenes are missing from evaluation data. Include construction changes and local traffic rules.
- Test simulation's contribution. Compare scenarios added through simulation with road, OEM and licensed data. Measure whether they expose failures or improve performance on held-out cases.
- Separate forecasts from commitments. When discussing eyes-off timing, distinguish an industry expectation from a product date. Ask what evidence must be met before making a customer promise.
What to watch
- Whether Basso's forecast of a major shift in 2027 into 2028 becomes a defined L3 availability milestone.
- Whether Wayve reports measurable safety results and thresholds for its four validation dimensions.
- How much simulation contributes to Wayve's data mix over time.
Send this to the autonomy and safety-validation leads setting next year's evidence plan.
Watch the minute where Basso gives the 2027 into 2028 forecast
Written by Real Alpha from the episode's transcript and the other moments it links, then checked against the transcript. Quotes are the speaker's exact words.
Minute by minute
A Pro write-up, open to you here. Try Pro free for 7 days
A Pro write-up. The minute by minute is in Real Alpha Pro; the episode, its moments and the transcript stay free.
Read it with Pro, 7 days free- 0:51Wayve targets autonomy across vehicles
Basso says Wayve aims to bring autonomy to all vehicles, not just robotaxis. He describes a path from L2/L2+ driver assistance to L4 robotaxis and says Wayve has a supervised public pilot in London. He frames the challenge as both building a capable AI driver and proving that it works.
- 2:45Safety needs evidence, not claims
Basso says driving remains risky and costly, making safety improvements a responsibility for engineers and product teams. For end-to-end systems, he says the industry has not yet established rigorous ways to prove they work, much as frontier AI developers face the problem of validating model outputs.
- 4:13Safety has four dimensions
Basso says Wayve assesses safety for people inside and outside the vehicle, comfort, compliance with local traffic rules, and competence in tasks such as lane changes. He says there is no single metric or benchmark that captures these qualities, so measures should reflect human expectations.
- 5:54Physical AI lacks a broad dataset
Basso says Wayve recognized driving and validation as data problems early and gathered varied data, including through a roadshow in very cold regions. He argues that physical AI lacks a general dataset comparable to language, so driving systems need diverse examples to improve.
- 8:02Driving behavior depends on context
Basso uses stoplights and construction scenes to show why fixed rules alone cannot capture driving behavior: signs and stopping conventions vary by place, and construction requires interpreting signs and their implications. He argues that learning from situations in different countries can help produce a driver that works worldwide.
- 11:16Wayve draws on several data sources
Basso says Wayve's operations team collects geographic and vehicle diversity as well as expert-driver reference data, while the company aims to keep its internal operations footprint limited. He also lists OEM partnerships and purchased or licensed sources, including dashcam footage. He says simulation should generate more data over time.
- 12:42End-to-end AI learns safety intrinsically
Basso says modular and hybrid systems can rely on maps and other checkers that may fail when signs move or appear temporarily, and he calls such checks a potential performance bottleneck. Wayve trains a safety representation to predict where driving will be unsafe, such as when a bus makes a lane change risky, and uses it to guide driving trajectories.
- 15:48L3 autonomy could give time back
Basso expects a major shift in 2027 into 2028 from L2+ hands-off assistance toward L3 eyes-off systems, where the vehicle is responsible for safety while occupants do other tasks. He says those systems need more stringent safety guarantees, but could let people use commute time for work, sleep or other activities.
- 18:26Wayve plans to make its case
Basso says Wayve's next steps include getting its system out, explaining its approach and building discussion with the public and customers about how it will prove safety. He describes that work as a way to move the industry forward.
Published Nov 8, 2026.
Transcript · 17 min read
MTS, hosted by Sophia Dew, with Brandon Basso.
Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.
And I think there's going to be a big shift in 2027 into 2028 when we move from. So look, Wave is working on these systems that are these L2 plus driver assist systems where you can take your hands off the vehicle. What we need to do is move beyond L2 plus to L3 systems where sometimes those are referred to as not just hands off, but eyes off systems. So those are systems where you can truly check out, do other tasks, and the system is responsible for safety. Hello, everyone, and welcome back to MTS. Today, I am joined by Brandon Basso, who runs the simulation and validation at Wave, the company teaching cars to drive with end-to-end AI. So Brandon, welcome to MTS. Thank you so much for having me. Thank you for joining. This is pretty exciting. I know we often, we're living here in San Francisco, we see things like Waymos and Zooks and all of these existing autonomous cars. But what you guys are doing is you're teaching new cars, how they can, too, be anonymous. Tell us a big picture for people who aren't familiar with Wave. Yeah. So like you mentioned, we're in San Francisco and it's hard to not run into an autonomous car here, whether it's a Waymo or a car from Zoox or any of the many other providers
out there. But there's also every other car on the road that isn't yet autonomous. And Wave's mission is to have ubiquitous autonomy everywhere for everyone. So all vehicles, not just Tesla's, every other vehicle out there. And this starts with a level of autonomy that we call L2 or L2 plus, which is sort of like, you can take your hands off the steering wheel, it can drive you in urban settings, highway settings, and then we're working on all the way up to L4 Robotaxi, where we have a public pilot in London, currently supervised pilot. What is the hardest part about this? Oh, my gosh, so many things. Well, I am the safety validation person. So for job security the reasons i have to say that part but um there are so many hard aspects i mean as an industry we've been at this problem for two decades going back to foundational research happening at berkeley and stanford and many other institutions around the world and there's so many hard parts i would broadly split it into two big categories and they're very much addressed in some of the content that we put out today so one is how do we make a driver drive better than a human superhuman an ai driver? And then two, how do we validate that it's doing what it's doing? Because it's of course
not good enough to just say, well, we built this great model and here you go. We have to prove that it works. And both of those parts have their own challenges. I personally focus a lot on the proof part, but they're both extremely hard when you think of the breadth of driving that can happen,
Read the whole transcript
“There's no one metric just in the way that you go to any AI leaderboard today.”
when we say what it means for a driver to be safe. Yeah, something you alluded to at the start of our conversation is you mentioned a tough piece is gathering enough data to train these cars to drive autonomously. Famously, even a decade ago, I remember I grew up here in the Bay Area, and you would see the people inside the Waymo cars just gathering data about what's happening on the streets. Where are we now? What's different about 2026 about how we gather all the simulation data? Yeah, absolutely.
So one of the core advantages and I would say core aspects of WAVE's approach going back to a decade ago is we very much recognized both the driving problem and the validation problem as a data problem. And we secured all sorts of interesting diverse data early on. We did this roadshow campaign where we drove well north of, I forget if it's the Arctic Circle or not, sorry, I'm not a geography expert, where it was very cold. And collected diverse data. and that is so critical because our system is fundamentally one that gets better with data just like any frontier llm today and people in robotics general robotics will tell you this there is no general data set for robotics today for physical ai there's no general data set like there is for language you know humans have developed all this language data throughout the years books and you know videos and stuff and we don't have um a corpus yet of um driving data so we work very hard on collecting that that's both our first party collection purchasing data going back to 10 years ago and then partnering with our fantastic oem partners and working with them on getting the data so that we can deliver back to them the best um the best products i want to shout
out our sponsor lovable you know the app you've been meaning to build the internal tool side project or product you'd otherwise lose a weekend to, with Lovable, you get software you can ship today. That includes an editable code base you can inspect and change, plus two-way GitHub sync. Lovable also handles the backend and infrastructure, including managed Postgres, auth, storage, hosting, payments, and more, so you don't have to wire it all up yourself. And through Lovable's MCP server, you can create and deploy your project directly from the agents you already use. Turn ideas into software people love with Lovable. Lovable.dev. Now, on to the episode. So I know today, during our interview, you brought along a demo.
Tell us a little bit about what this demo is supposed to be.
“if we can get that and we can incorporate something that happened on a highway 10 years ago into our corpus, then we can build a better driver.”
So multi-prong. All right. With that said, I do want to see the other demos that you have on here. Yes. So we kind of jumped into the part that's near and dear to my heart, which is the validation
“the main message here is that the way that we've built the driving stack is with safety as an intrinsic property versus an extrinsic bolt on property”
“I think there's going to be a big shift in 2027 into 2028”
get the system out there autonomy for everywhere for everyone it's to tell our story it's to represent why we think that this is the right way to solve the problem it's it's honestly to do posts like this and talks like this to sort of um move the industry forward um and um be a thought leader in the space uh both with regard to the safety of the driver itself but also how we plan on doing validation um so that we can start having a conversation with the public and with our customers about how we plan on proving that these systems are going to make our lives better. Very exciting. Brandon, really wanted to thank you for joining us today on MTS and for bringing your demo as well. We got to show it. The whole audience got to see what it actually looks like.
I am so excited for the future of autonomous driving to have a world where people can get their time back, where transportation becomes a lot easier and more efficient. Very exciting. And this is at the frontier of what is possible. So thank you so much for joining us today on MTS. A huge thanks to MTS sponsors. Credits for your startup at neon.com slash MTS. Special thanks to our sponsor, Kong, the AI connectivity platform. Connect APIs, LLMs, agents, and systems with serious security and governance. KongHQ.com. Finally prove what your AI investment is worth with Lairdyn. Head to Lairdyn.com today to start measuring your AI's bottom line.
MongoDB, trusted to run critical applications by 75% of the Fortune 100, The intelligent data platform for the era. Apply today for access to up to $15,000 in Atlas credits, dedicated technical support, and exclusive events and go-to-market opportunities. MongoDB.com Chargebee. Billing and monetization infrastructure for the AI era. Chargebee.com Blitzy. Autonomous software development for enterprise code bases. Ship 5x faster. Blitzy.com AdQuick. Make your brand a billboard. out-of-home advertising as easy to scale as digital adquick.com arena measuring ai performance in the real world arena.ai support for the show comes from vcx the public ticker for private tech the u.s stock market started history's greatest wave of wealth creation from factory workers in detroit to farmers in omaha anyone could own a piece of the great american companies but today our most innovative companies are staying private longer which means everyday americans are missing out. Until now. Introducing VCX, a public ticker for private tech. Visit getvcx.com for more info.
That's getvcx.com. Carefully consider the investment material before investing, including objectives, risks, charges, and expenses. This and other information can be found in the Innovations Fund Prospectus at getvcx.com. This is a paid sponsorship. You
