What's up everyone and welcome to Next in Dev, a podcast where I cover all the news and topics that have happened recently in modern web development.
In today's episode I'm going to cover if ChatGPT 5 worth the hype, what really landed in Next.js 15.5, and all of the quick wins that are hiding in Payload, Figma and more.
We'll start with payload, which just had two new versions come out in recent weeks versions 3.51 and version 3.52.
These versions were pretty light on features and had a ton of bug fixes, so I encourage you to look through the release notes to see if any of those bug fixes apply to you.
One feature is that tenant field overrides is now available in the multi-tenant plugin.
Now you can override most of the tenant fields using the tenant field override option in the plugin.
Also in the multi-tenant plug in the tenant selector label is being deprecated in favor of exposing the entire internationalization object in the config.
There's also been an update to the UX for the query selector presets.
The Group by feature, introduced in 3.49, now lets you group by boolean fields, and if dates are null.
In version 3.52, there's now a page and limit option for you in the Import and Export plugin.
Figma, which is Payload’s parent company, released ten new scatter brushes to their Figma draw product.
These brushes help add more interesting shadows and stippling effects to your designs in Figma Draw.
Figma also had some updates to their AI image tools, which now allow you to bulk remove backgrounds for up to 25 images and also boost resolution using their AI tools.
You can now switch between Gemini 2.0 flash or GPT image one in your model selector in Figma.
Next.js continues to have almost daily, sometimes more than daily, commits to their canary branch.
but the big thing for Next.js is that 15.5 has now been released.
now there are a ton of release notes here.
And I really do encourage you to go read it because I'm not going to read it.
So if there's anything in there that you might need to see, you'll want to go in there and actually read the release notes to make sure that you are up to date with everything you need for Next.js 15.5.
But to sum up what I could see, there were some caching mechanism updates, some react version updates, as well as some work done to turbo pack.
In less flashy news, Next.js version 15.4.6 also came out to backport some bug fixes to version 15.4.
The canary branches seem to be primarily focused on turbo pack and caching, but I did see a couple of commits that are working on partial pre rendering as well as more robust error messages.
V0.dev is now v0.app.
And now I'm sure there's some hip, cool, marketing reason why they changed the name, but to me everything seems to be working about the same as it did when it was under V0.dev.
And despite the vibe code-y reasons, I'm sure it doesn't really matter too much in the long run, but I am curious when they are going to try and purchase a v0.com and how much that's going to cost them.
and while I was writing the script, I actually checked if v0.com was available.
It's not.
Vercel now claims that V0 now uses agentic AI to run code, debug code, generate code, and plan the code and all of that.
But if I'm honest, I ran into a lot of the same issues as I did before using v0.dev that I was running into with v0 dot app.
So for me it's not a huge improvement.
But tell me in the comments.
Maybe you're seeing something a little bit different than I am.
Vercel, the maintainers of Next.js and Shadcn, now allow for GPT 5 to be used in their AI gateway.
I'll discuss open AI and ChatGPT 5 a little bit later on, but Vercel is clearly moving just as fast as these AI companies, who seem to be releasing new models every other week.
In fact, in more AI support news.
Claude sonnet 4’s 1 million token context window is also now supported in Vercel’s AI gateway.
If you have any Vercel deployments and you're like me, you may have been seeing some notifications that NodeJS 18 is being deprecated from Vercel’s platform.
Now Vercel is trying to make it easier on us to update all of these projects that we're working on by providing a bulk editing tool to upgrade Node.js for you.
They're trying to make it really easy to upgrade, and I still haven't done it.
There have been several updates to Vercel’s invisible CAPTCHA, BotID, which is meant to catch pretty sophisticated bots but not add any friction to real users.
One update is that they're starting to use a verified bot directory, similar to what Cloudflare has been using.
Another update is that the Vercel team has improved their fake hardware detection.
In big caching news you can now use Vercel’s runtime caching API.
this appears to be the next iteration of how Next.js handles caching.
As if it wasn't convoluted enough.
Now, before we move on, if you're enjoying this video and finding it useful, please take a second to make sure that you like this video and to check that you're subscribed to my channel for more updates like this, you can now find my Next in Dev videos like this one on Spotify and Apple Podcasts, so follow there if you prefer to get your podcasts in that way.
Also consider becoming a member of my YouTube channel to get access to exclusive perks like early access to content and member shout outs, polls, posts, and more.
There's a link to sign up in the description.
Thank you for your support.
Now let's get back to it with Tailwind.
Tailwind has actually released something this week.
It's been about a month, maybe five weeks since tailwind has released anything.
But I still don't really have much to say about tailwind because all they did in version 4.1.12 was some bug fixes.
And not even like good bug fixes or common bug fixes, just general maintenance bug fixes.
So we're going to move on.
But do know that there was a new tailwind release.
And if you're interested, you can look at the release notes.
Now in AI news Anthropic, a couple of weeks ago released a new statement on their usage policy that they claimed only affects about 5% of their user base.
The usage limits are now set weekly and if I'm honest, haven't affected me in the slightest.
But I do wonder who they have affected.
I've seen a couple of posts on Reddit asking that same question about who's using Claude Code essentially 24/7 to even get close to these weekly limits.
If that's you.
If you're feeling the squeeze and I'm just not seeing it, let me know in the comments.
I'm really interested to know who that 5% of Anthropic’s user base really is.
In fact, I know a few power users who have reported no real effect to them since the usage policy went into effect.
Claude Opus 4.1 was also announced.
Based on software engineering benchmarks, this was a modest, marginal increase, yet still significant.
But it does appear that we're starting to reach that point of marginal returns.
Anthropic is also addressing their tool's part in cybersecurity, politics, and high risk questions like legal and financial questions, as well as their tool’s usage by law enforcement.
In lighter news, as something that I mentioned in passing earlier on Claude Sonnet 4 can now accept up to 1 million tokens in context, which equates to about 75,000 lines of code to be processed all at once.
Now you can process up to 75,000 lines of code.
But whether or not you should is only a question that you can answer.
ChatGPT 5 was released with great fanfare this past week, and if we're honest, it was kind of a flop.
Their goal, as they stated, was to replace all of their GPT 4 models with one unified GPT 5 model.
But that didn't last.
OpenAI brought back their GPT 4 and other legacy models pretty quickly after GPT 5 was meant to replace them.
outside of more outlandish comments from Sam Altman about how AI is soon going to out talk humanity, there's not really a whole lot going on with OpenAI.
Which brings me back to my point with Anthropic and Claude Opus 4.1, where I'm now starting to see marginal returns from these new releases.
GPT 5 was a flop.
Not because it's not a good tool, but because it's not a much better tool than GPT 4 was.
And if you're starting to notice a trend in all of this AI news, we just get comments from AI CEOs to pay more attention to their company.
So when Sam Altman says something about how AI is going to take your job, how AI is going to take over the world, how all of our jobs are going to be done through prompt commands and text commands.
You have to remember that these people are just trying to get more attention on their company.
So take everything they say with a grain of salt.
Now, whether it's true that we'll see some job loss because of AI or some shifting in how work happens because of AI, that's a different debate.
But just because Sam Altman, Elon Musk and all of these other AI company heads say something does not necessarily make it true.
many of you are using or considering using Cursor Well cursor is now supported on Vercel’s MCP.
This allows you to interact with your projects and fetch logs without ever leaving cursor.
It's really simple to set up, and it's really making me want to try cursor even more.
The cursor agent is now available from the CLI.
This means you can use the cursor agent even in any other IDE, whether it's neovim or JetBrains or any other one.
so you can get the full power of the cursor agent without even using cursor.
GPT 5 is also now available in cursor.
Again, this announcement came with more hype about the model that seems to have largely gone unproven, but the fact remains that very quickly after GPT 5 released, you were able to use it in cursor.
And now I'm sure Gemini is great, but I've spent a large portion of this video talking about AI, so I really want to move on to Railway The big news here is that you can now apply templates to existing projects.
Before you had to create a new project and then use a template.
But that doesn't really help you if you have an already existing project and you want to add something to it.
So for example, if you wanted to host plausible on Railway and you wanted to use a template that you found, you wouldn't be able to do that in your current web project, you would have to create a new project and then have those projects communicate with one another.
So now you can just add that plausible template into your current web app, and railway will take care of the rest after you get everything configured.
Now that all of this is done and available, it gives a little bit more incentive to use Railway’s kick back program for new templates where you get 25% of the cut of whatever Railway makes from people using your templates.
You now need to verify your billing information for Railway so they can start collecting the appropriate taxes for your country, whether sales tax or VAT.
You'll need to update this information in your Railway account by September 1st.
More on AI, Railway has officially released their MCP server.
You can now automate deploying and managing your apps through Railway’s MCP This acts pretty similarly to how Vercel has theirs setup, but it's still early and experimental, so feel free to play around with it.
But maybe don't rely on it quite yet.
and lastly Railway has released improvements and a commitment to continue improving their databases.
And that's all I have.
But I know that's not all there is.
So tell me what I missed.
Leave a comment that has a suggestion or a topic that I should cover, or something that you thought was big in the news that should have been discussed, and you would like to discuss it further in the comments.
Again, don't forget to become a member of my YouTube channel.
You'll receive exclusive perks like early access to content, and there's a way to sign up to be a member of my channel in the description Like the video and check to make sure you're subscribed if you're watching on YouTube.
And you can also follow and rate the show if you're listening or watching on Spotify or Apple Podcasts.
That's all I've got.
Thanks for watching and I'll see you on the next one.
