sparta

A School District's Effort to Educate Waymos to Halt for School Buses Did Not Succeed

A School District’s Effort to Educate Waymos to Halt for School Buses Did Not Succeed

sparta

The claimed advantage of autonomous vehicle technology lies in the ability of each car to learn from the errors of others. As detailed on Waymo’s website: “The Waymo Driver benefits from the shared experiences collected across our fleet, including earlier hardware versions.”

Nevertheless, in Austin, Waymo’s vehicles struggled for several months to appropriately stop for school buses while picking up and dropping off students. An official from the Austin Independent School District (AISD) asserted that in at least 19 instances, the vehicles “illegally and dangerously” bypassed the district’s school buses while red lights were flashing and stop arms were extended, failing to make complete stops as mandated by law.

In early December, Waymo even launched a federal recall concerning these incidents, acknowledging at least 12 of these to the National Highway Traffic Safety Administration (NHTSA), which oversees road safety. Federal paperwork indicates that the self-driving vehicle company’s engineers had “created software modifications to rectify the behavior” weeks earlier.

However, even after the recall, reports of school bus violations continued, as noted by school officials and a report from the National Transportation Safety Board (NTSB), an independent federal safety organization that is also looking into the issue.

Communications via email and text between school officials and Waymo, acquired by WIRED through a public records request, highlight the measures taken by the Austin public school district and Waymo to address the problem. AISD held a half-day “data collection” event in a school parking lot in mid-December, as the records indicate, where several staff members set up school buses and stop-arm signals for the self-driving company to collect data on the vehicles and their flashing lights.

By mid-January, one month later, the school district reported at least four additional instances of school buses being passed in Austin.

Sora's closure may serve as a wake-up call for AI video.

Sora’s closure may serve as a wake-up call for AI video.

sparta

This week, OpenAI revealed that it is discontinuing its Sora application and associated video models merely six months post-launch.

During the most recent episode of TechCrunch’s Equity podcast, Kirsten Korosec, Sean O’Kane, and I discussed the implications of this decision for OpenAI and the broader industry. To a degree, this action aligns with what we’ve been observing regarding OpenAI’s pivot towards enterprise and productivity solutions as it moves closer to a potential IPO.

In fact, Kirsten noted that OpenAI’s choice to close Sora reflected “a maturity level that was encouraging to witness in an AI lab.”

However, the termination of Sora—alongside ByteDance’s indicated postponement in the global rollout of its Seedance 2.0 video model—might serve as a wake-up call for creators of AI video technologies, and for advocates who assert these tools will imminently usurp Hollywood.

Below is a preview of our discussion, shortened for conciseness and clarity.

Anthony: It’s important to point out that it’s not merely the app. Personally, I found the app notably unappealing, as did others, due to its premise of a social network devoid of actual users, filled only with irrelevant content.

Beyond the app, it appears OpenAI is essentially scaling back nearly all its video-related activities. The Wall Street Journal, which initially reported some of this information, suggests that OpenAI is prioritizing business-centric products and enterprise solutions ahead of a possible public offering, indicating that this consumer-focused social app—and video endeavors in general—are not currently key priorities.

Techcrunch event

San Francisco, CA
|
October 13-15, 2026

Sean: I never really engaged with [the app]. The concept itself didn’t resonate with me for several reasons. And, you know, this serves as a good reminder that OpenAI—without intending to disparage them in any way—probably recognized, internally, the element of chance involved in ChatGPT’s success.

Clearly, there is something valuable for users, and I don’t want to diminish that, because achieving the reported usage statistics is indicative of a genuinely working model—and even more so that it has been maintained and evolved into something that remains relevant for users over time.

However, when Sora launched, it exuded an air of, “We’ve created the most successful consumer product ever, and now we’re duplicating that success with this. We’re going to include Disney and all this other content.” This serves as a stark reminder that there isn’t always a guaranteed shortcut to creating the greatest consumer products, and there must be intrinsic value for them to endure.

Kirsten: I actually commend OpenAI for this choice, as we often jest about the concept of “move fast and break things,” but there is merit in companies that can rapidly iterate and discontinue underperforming products without feeling a sense of defeat. There was significant financial investment involved; the arrangement with Disney amounted to a billion dollars. However, without access to their financial statements, we can only speculate about their expenditures and any long-term value.

While it was intriguing to explore the possibilities they could produce, their decision to shut it down, in my view, reflected a maturity level that was positive to see from an AI lab.

Anthony: In terms of implications for OpenAI, this action appears to align well with the strategy we’ve been hearing regarding their future direction. It doesn’t seem like a severe setback for how we perceive the future of generative AI.

Especially in the video realm, it is notable that this coincides with recent developments surrounding Seedance, ByteDance’s generative AI model for video. Reports indicate that [Seedance 2.0 has] been delayed due to engineering and legal considerations, specifically regarding “Can we integrate IP protections into this?” which seems to have been an overlooked issue previously.

Thus, we find ourselves at a crucial reality check moment. There were numerous exaggerated claims, including those from Hollywood insiders proclaiming, “This is it, we’re done—just input prompts and create films.” Reality has shown that, for numerous technical and legal reasons, the path to that outcome is far more complex and distant.

Sean: Lastly, it’s worth mentioning that this decision is one of several occurring since Fidji Simo began overseeing daily operations. This has brought a significant shift within OpenAI. As we move further away from her taking the helm for these consumer products and determining their fates, it will likely become easier to retrospectively assess how pivotal this moment was for the company.

Loading the player…