WEBVTT

1
00:00:00.380 --> 00:00:09.840
What's up everyone? Welcome to Next in Dev, a weekly overview of all the news I could find in the modern web dev industry. This week was rough for transparency in the AI industry.

2
00:00:10.100 --> 00:00:20.560
OpenAI published a safety blog post for Sora and then killed the entire product a day later. Cursor launched a, quote, "frontier-level coding model" and forgot to mention it was built on someone else's work.

3
00:00:20.600 --> 00:00:27.170
And a federal judge told the Pentagon its Anthropic ban looks less like a national security issue and more like punishment.

4
00:00:27.200 --> 00:00:38.120
Earlier this week, OpenAI published a detailed blog post about Sora safety, including consent-based likeness controls, metadata on every video, layered content filters, teen protections, and more.

5
00:00:38.180 --> 00:00:48.670
It read like a company that was about to settle in for the long haul with its AI video platform. The next day, OpenAI killed Sora. The standalone app, the API, and the website are all shutting down.

6
00:00:48.680 --> 00:00:59.920
Publishing a safety blog post the day before you kill the product is peak OpenAI. According to Reuters, Disney teams were actively collaborating with OpenAI on a Sora project Monday evening.

7
00:00:59.960 --> 00:01:07.160
30 minutes after that meeting wrapped, Disney found out Sora was being canceled. One person involved called it, quote, "A big rug pull."

8
00:01:07.180 --> 00:01:16.560
The $1 billion deal Disney announced just three months ago, which licensed over 200 characters and allowed Disney to take a stake in OpenAI, never actually closed.

9
00:01:17.020 --> 00:01:29.020
No money ever changed hands, and Disney is now exiting the partnership entirely. Some Sora staffers were themselves blindsided. So why is OpenAI doing this? Well, downloads dropped 45% from their November peak.

10
00:01:29.200 --> 00:01:37.000
Total in-app revenue amounted to roughly $2.1 million, which, let's be honest, is a rounding error for a $730 billion company.

11
00:01:37.220 --> 00:01:44.140
OpenAI says the compute needs to go elsewhere, and internally they're consolidating around a ChatGPT super app strategy.

12
00:01:44.400 --> 00:01:53.560
Multiple outlets explicitly cite Anthropic's Claude Code gaining developer traction as a competitive pressure, pushing OpenAI toward enterprise and coding tools.

13
00:01:53.640 --> 00:02:03.940
OpenAI is conceding ground in creative tools to fight on Anthropic's turf, and that's a significant pivot for a company that spent two years positioning AI video as a flagship capability.

14
00:02:04.120 --> 00:02:12.400
My opinion is that we'll see more practical uses of AI take center stage because of this. Generating video and audio is fun, but there's not much use to it.

15
00:02:12.580 --> 00:02:23.760
Focusing on enterprise functions and coding are much more valuable uses of AI, so the shift makes sense. OpenAI's foray into AI-generated creativity definitely has them behind the ball though.

16
00:02:23.800 --> 00:02:31.080
Of course, OpenAI isn't the only company this week with a gap between what it said publicly and what was actually happening behind the scenes.

17
00:02:31.340 --> 00:02:42.640
Cursor released Composer 2 last week, positioning it as their own, quote, "frontier-level coding model". It had strong benchmarks and is priced aggressively at 50 cents per million input tokens.

18
00:02:42.680 --> 00:02:49.329
The blog post credits their first continued pre-training run and reinforcement learning on long-running coding tasks.

19
00:02:49.489 --> 00:02:58.740
Within hours of launch, a developer intercepted Cursor's API traffic and found the model identifier in plain text. Surprise, it was Qimi all along.

20
00:02:58.760 --> 00:03:06.420
Composer 2 is built on Qimi K25, an open source model from Moonshot AI, which is a Chinese startup backed by Alibaba.

21
00:03:06.600 --> 00:03:18.580
Cursor confirmed it the same day, and Cursor's co-founder acknowledged it was, quote, "a miss" not to disclose it. Moonshot later confirmed the arrangement was an authorized commercial partnership through Fireworks AI.

22
00:03:18.860 --> 00:03:23.920
Building on open source models isn't scandalous. It's how a lot of AI development works now.

23
00:03:23.940 --> 00:03:31.040
But saying, quote, "our first continued pre-training run" is technically accurate and strategically misleading at the same time.

24
00:03:31.049 --> 00:03:42.430
The engineering Cursor did on top of Qimi is real, and that makes the lack of transparency even more confusing. If your added value is strong enough to stand on its own, why obscure the foundation?

25
00:03:42.660 --> 00:03:54.880
It gets thornier when you look at licensing. Qimi K25's modified MIT license requires prominent attribution above $20 million in monthly revenue. Cursor reportedly clears $160 million a month.

26
00:03:55.040 --> 00:04:03.520
Their interface showed Composer 2 with no mention of Qimi. And the headline benchmark they lead with, CursorBench, their own proprietary benchmark.

27
00:04:03.840 --> 00:04:12.172
They include third-party results too, but grading your own homework on a model already under scrutiny for origination-Is a pattern worth paying attention to?

28
00:04:12.392 --> 00:04:19.832
Attribution in the open source AI ecosystem needs to be a norm, not an afterthought. It should not be something so easily missed.

29
00:04:20.112 --> 00:04:28.952
A federal judge in San Francisco said this week that the Pentagon's ban on Anthropic looks like an attempt to punish the company for publicly criticizing the government.

30
00:04:29.012 --> 00:04:36.611
If you're not aware, Anthropic's CEO said in February that Claude wouldn't be used for autonomous weapons or mass surveillance of American citizens.

31
00:04:36.752 --> 00:04:41.612
The Trump administration wanted Claude available for, quote, "all lawful purposes".

32
00:04:41.652 --> 00:04:49.402
When they couldn't agree, the Pentagon designated Anthropic a, quote, "supply chain risk" and Trump ordered all agencies to stop using Anthropic products.

33
00:04:49.672 --> 00:05:01.452
At a hearing on Anthropic's request for a preliminary injunction, a judge called the government's actions, quote, "troubling", saying they, quote, "don't really seem to be tailored to the stated national security concern".

34
00:05:01.692 --> 00:05:13.792
If the worry is operational integrity, the Pentagon could just stop using Claude. Instead, the Pentagon announced that any contractor doing business with the military must cut all ties with Anthropic.

35
00:05:13.952 --> 00:05:22.012
One brief described it as, quote, "attempted corporate murder". The judge didn't go that far, but said it looks like an attempt to cripple Anthropic.

36
00:05:22.222 --> 00:05:31.992
The government's lawyer conceded during the hearing that the designation doesn't actually stop military contractors from using Anthropic on non-military work.

37
00:05:32.312 --> 00:05:43.732
This walks back on previous broader statements and should assuage the fears of some enterprises that were on the fence about if they were able to use Claude or not. This case matters well beyond this one company.

38
00:05:43.772 --> 00:05:52.492
The legal question is whether the government can repurpose supply chain risk designations against a domestic company for setting terms on its own product.

39
00:05:52.552 --> 00:06:00.312
The outcome sets a precedent for every AI company about what happens when you say no to the Pentagon. A ruling is expected in a few days.

40
00:06:00.352 --> 00:06:11.892
Taking a page out of Vercel's playbook, GitHub announced that starting April 24th, interaction data from Copilot Free, Pro, and Pro Plus users will be used to train AI models unless you opt out.

41
00:06:11.932 --> 00:06:22.272
That includes your inputs, outputs, code snippets, surrounding context, file names, repository structure, and your feedback on suggestions. Business and enterprise users are not affected.

42
00:06:22.492 --> 00:06:32.072
If you previously opted out of the product improvement data collection setting, your preference carries over. But if you never touched that setting, you're now opted in by default.

43
00:06:32.132 --> 00:06:44.032
GitHub says incorporating real-world interaction data from Microsoft employees has already improved acceptance rates, and broadening that to all users will make models better at catching bugs and understanding workflows.

44
00:06:44.212 --> 00:06:54.612
Your next step, probably to go to your GitHub settings under privacy and check your Copilot data preferences before April 24th. After that date, opting out only prevents future use.

45
00:06:54.652 --> 00:07:06.292
It won't retroactively remove data already collected. Given that this comes just days after Vercel's own opt-in by default terms of service update for free tier users, there's a clear pattern forming.

46
00:07:06.492 --> 00:07:18.772
If you're on a free or low-cost tier, your usage data is the product and will be used to feed AI. Next.js just addressed the single biggest criticism it's faced for years: platform lock-in.

47
00:07:19.012 --> 00:07:28.892
The Adapter API, which went stable in 16.2, gives any hosting provider a typed, versioned contract they can build against to deploy Next.js apps at full fidelity.

48
00:07:29.112 --> 00:07:37.552
Streaming, server components, partial pre-rendering, middleware, caching, and more are now formally documented and available through a public API.

49
00:07:37.792 --> 00:07:48.542
You can bet that Cloudflare forking Next pushed Vercel over the edge on this one. Vercel's own adapter uses this same public contract with no private hooks or special integration path. It's open source.

50
00:07:48.772 --> 00:08:01.072
Any platform that passes the shared test suite can become a verified adapter hosted under the Next.js GitHub organization. Adapters for Netlify, Cloudflare, and AWS through OpenNext are in active development.

51
00:08:01.172 --> 00:08:04.652
A Bun adapter already exists as a reference implementation.

52
00:08:04.712 --> 00:08:15.952
This happened through direct collaboration with OpenNext, Netlify, Cloudflare, AWS Amplify, and Google Cloud, formalized through a new ecosystem working group with public meeting notes.

53
00:08:16.162 --> 00:08:22.042
Their announcement post includes quotes from engineers at Netlify and Cloudflare who were a part of the process.

54
00:08:22.272 --> 00:08:29.251
For years, the argument against Next.js was that Vercel controlled both the framework and the only first-class deployment platform.

55
00:08:29.472 --> 00:08:38.862
This release doesn't completely resolve that tension because Vercel still has a head start on optimization and will always ship adapter support for new features first.

56
00:08:39.072 --> 00:08:50.162
But it fundamentally changes the relationship between Next.js and the rest of the hosting ecosystem. If you've been hesitant about Next.js because of the Vercel dependency, this is worth a serious second look.

57
00:08:50.492 --> 00:09:01.612
Figma opened its canvas to AI agents this week through a new use Figma tool on their MCP server. Agents can now create and modify design assets using your existing components, variables, and tokens.

58
00:09:01.772 --> 00:09:07.132
This allows agents to read your library first, so output reflects your design system instead of looking generic.

59
00:09:07.312 --> 00:09:17.732
The standout addition is skills, markdown files that define how agents should work in Figma, including conventions, sequencing, and what good looks like. Anyone can author one without writing any code.

60
00:09:18.172 --> 00:09:23.912
It's free during beta while Figma figures out pricing, and it works with Claude Code, Codex, Cursor, and Copilot.

61
00:09:24.122 --> 00:09:34.852
The skills system means the more you invest in your design system, the more value you get from your agent workflows, but the pricing model is the thing to watch. Now for a few rapid-fire updates.

62
00:09:34.992 --> 00:09:45.522
Payload version 380 added multi-tenant slug uniqueness control, plus fixes for Drizzle pagination, Postgres Near queries, SQLite scheduled publish, and lexical RTL support.

63
00:09:45.752 --> 00:09:54.512
WebStorm 2026.1 enabled the service-powered TypeScript engine by default and added multi-agent AI chat via the agent client protocol.

64
00:09:54.692 --> 00:10:05.412
TanStack DB version 0.6 added SQLite-backed persistence across browser, React Native, Electron, Tauri, and durable objects, plus hierarchical data projection with includes.

65
00:10:05.552 --> 00:10:18.472
TanStack also removed all third-party ads from TanStack.com. Claude Code pushed versions 2.1.79 through.84 with transcript search, bear mode for scripting, rate limit display, and PowerShell tool preview.

66
00:10:18.832 --> 00:10:29.632
Railway redesigned its dashboard, added DNS management and raw queries for all database types, and rolled out Claude-powered deploy diagnostics. What did I miss?

67
00:10:29.672 --> 00:10:33.772
There's so much happening in modern web dev that I know I missed something.

68
00:10:33.972 --> 00:10:44.992
Let me know by leaving a comment wherever you're watching or listening, or by joining my Discord server and subscribing to the Next in Dev newsletter at nlvcodes.com. Thanks for watching or listening.

69
00:10:45.212 --> 00:10:46.192
See you in the next video.

70
00:10:53.432 --> 00:11:06.492
[outro music]