Sep 21 · AI Daily

Nissan CEO Ivan Espinosa revealed that the 14th-generation Skyline (V38) sedan will officially debut this December. The project took just 26 months from kickoff to completion, more than half the 55-month timeline of the previous V37 generation, thanks to the widespread use of AI across design, engineering, testing, and production. The new car uses a rear-wheel drive and FM front-midship platform, is expected to offer 6-speed manual and 9-speed automatic transmissions, features a 3.0-liter turbocharged V6 engine, and is evaluating a hybrid version.

A product lead shared a six-step framework they follow for handling almost any problem: understand existing information, fill in missing information, define the problem, define a clear and simple solution, define goals, act with urgency, and loop back to steps 3-5 repeatedly as new information arrives. They noted the most common failure patterns are failing to clearly define the problem and failing to provide a clear, simple solution, adding "I like making mistakes" because mistakes help clarify problems and find the correct solution faster.

An engineering lead at Imprint documented their practice of implementing a software factory pattern within the company. The core is the /linear-project-loop skill, which audits Linear project goal definitions (Notion RFCs, Datadog, or Snowflake metrics), automatically adds or removes issues after filling in missing tools, handles non-blocking tasks, and periodically reruns the loop.

The Flet team released Flet 1.0 and announced it is ready for production application development. The open-source Python framework renders UI based on Flutter, and flet build supports eight target platforms: apk, aab, ipa, ios-simulator, windows, macOS, Linux, and web. Apps can bundle Python 3.12, 3.13, or 3.14, and it is available on PyPI under the Apache 2.0 license.

Bezos pushed back against claims that AI will replace radiologists and software engineers, arguing that AI will enhance the capabilities of these practitioners. Comparing it to someone handing you a bulldozer while digging a basement, he said economic productivity will increase significantly and could instead result in a labor shortage.

Oops. Being "fast" used to be a bad thing 💅

Jev is now open to everyone with no waitlist. Get started here: https://console.typesafe.ai

Baseten CEO Tuhin Srivastava (@tuhinone) just dropped 2 numbers showing how fast AI inference is growing:

Token usage on Baseten is up 40x year-over-year, while revenue grew roughly 10x over the past 12 months.

From the "Sources Podcast and Alex Heath" (@alexeheath) YouTube channel (full video link in comments)

Are top Chinese AI labs using Nvidia chips for training and starting to use Huawei chips heavily for inference? Is that how it works?

An employee who joined a large company half a month ago leaked that the team's specs, code, tests, PRDs, tickets, and reports are all generated by Claude Code, with engineers from L1 to L7 doing the exact same thing. No one likes this approach, but senior management requires maximum output. Employees work 12 to 13 hours a day just to hit return, and no one reads any of the generated output.

Back in July, a report went viral online claiming that Waymo is more dangerous than for-hire vehicles in New York. That analysis was later proven flawed, and an updated report found Waymo to be much safer. Good thing they updated it with new data, though https://www.openplans.org/how-for-hire-autonomous-vehicles-impact-nyc

Jev is now open to everyone. No waitlist. Get started here: https://console.typesafe.ai

You cannot stop MUSE MADNESS

[Quoting @mignano]: Two things feel inevitable:

1. Muse
2. Muse having a major impact on Meta's trajectory

Our co-founder @aidangomez started Cohere to let you do more of what you want to do and automate what you need to do.

The AI we're building helps handle the tedious tasks, leaving you more time to focus on what truly drives you.

I don't completely agree with the premise that "full RSI is happening," but this is an exceptional video summarizing where we stand right now. https://youtu.be/fjZ90V_JREk?is=1jf1RRuWRTjo9_Iw

🔥 Introducing the Jev Primer and Playground

Want to understand how Jev works? Look no further.

This is the only interactive guide you need to get started.

No more confusion about what Jev can and cannot do. We also built a dedicated Jev Playground so you can try out various use cases right now.

Laya (OS Jev) runs offline on Mac M4 via CoreML, reaching 45 decisions per second. Users can install laya-coreml via uv [demo], download the aac6fef/laya-multilingual-coreml-ane model locally, and run the laya-coreml-snake demo.

Nvidia CEO Jensen Huang said in a CBS News interview that the chance of the world ending in 2030 is 0%, calling AI doomsday claims scientifically unfounded and irresponsibly alarmist. His remarks responded to former Anthropic researcher Alex Conxjo's view that developing superintelligence gambles with human lives. He believes existing U.S. product liability laws and cybersecurity laws are already sufficient, requiring no additional legislation to regulate AI developers, and opposed banning U.S. chip exports to China, saying every chip company should compete globally.

This opinion piece argues that frontier AI labs exaggerate the risk of rogue agents to lobby the U.S. government for antitrust exemptions and curb open-source competition.

Ethan Mollick advocates expanding this call from economics to the entire social sciences: the AI field changes too fast to wait decades for papers with perfect causal identification, requiring fast, forward-looking research deeply supported by AI capabilities. Citing @alexolegimas, he noted that while economists are accustomed to pursuing clean identification, the current moment calls for methodologically reliable, directionally diverse real-time signals, such as multiple independent teams reaching the similar conclusion that "hiring for entry-level positions with high AI exposure is slowing down."

Anthropic chief economist Peter McCrory responded on how to read papers on the economic impact of AI, proposing four practical suggestions: prioritize analysis regularly updated with new data, publish work publicly and revise views based on new evidence, acknowledge uncertainty and make falsifiable predictions, and maintain seriousness and humility.