Field Notes [07.2025]

Hi friends,

It’s been a minute! Always Scheming has been on pause over the last few months but we’re swinging back into action with a brand new dispatch called Field Notes.1

Field Notes is where I track signals at the edges of interactive entertainment and emerging tech. These won’t be the over-discussed headlines leading every gaming blog and podcast. Rather, Field Notes will center on subtle shifts, interesting prototypes, underreported headlines, and early eye-catching experiments.

Think of it as my working notebook: half radar, half research log, all curated for the curious and packaged up into a snackable monthly communiqué.

Let’s get into it.

Earlier this year, Bloomberg reported on a new pair of Meta smart glasses set to debut later this year. Codenamed “Hypernova,” this latest iteration is purported to include a HUD in one lens for displaying time, weather, turn-by-turn navigation, and other sorts of notifications. The hardware will also reportedly include a “neural wristband” that allows for finger gesture controls. All of this is due to land somewhere in the $1,000 - $1,400 price range.

This had all been previously reported, if underdiscussed. However, UploadVR is now reporting on new leaks out of Meta just this month related to the Hypernova project. It seems dataminers have uncovered early firmware for the glasses and surfaced detailed imagery, including one render that shows a potential product name: “Celeste.” We also got a closer look at the finger gestures, the wristband design, and the notification icons used in the HUD.

While none of this is entirely earth-shattering, we flagged it for two key reasons:

All this represents a firm incremental step forward from the success of the Meta Ray-Bans. However, it’s worth noting that Hypernova/Celeste will not actually be a next generation of the Meta Ray-Ban line. The owner of the Ray-Ban brand, EssilorLuxottica, reportedly balked at the design compromises necessitated by the HUD.2

Indeed, it’s fair to say that the new tech will come with drawbacks: the HUD requires a heavier pair of glasses (70 grams vs. the 50 gram Ray-Bans), while the new vector for interaction, finger gestures, comes with the added encumberment of a (questionably fashionable) wristband.

Presumably this will all improve with time and iteration — the aforementioned Bloomberg report mentions a follow-up model with a two lens HUD already targeted for 2027 — but the addition of the HUD is an important milestone as it enables a broader range of application development for smart glasses as a platform, including gaming.

As I previously wrote in Meta vs. Reality:

We’re certainly optimistic about the potential for AI and voice input to incubate interesting new interactive entertainment experiences, but the Meta Ray-Bans really ain’t it (at least, not in their current form).

Adding the HUD definitely opens up new possibilities for interactive entertainment, though I’m under no illusions that this will be a gaming-first (or second, or third…) piece of hardware. Nevertheless, the Meta leaks also revealed a built-in minigame titled “Hypertrail” that “may be inspired by Galaga…but also involve the user's location in some way.”

Worth keeping an eye on! But only one eye because, you know, one lens HUD and all…

Speaking of smart glasses applications, the other reason we flagged this story for Field Notes was the mention of mapping and turn-by-turn navigation.

Previous UploadVR reporting noted the availability of a Maps app on Hypernova prototypes and these latest leaks seem to confirm its inclusion. Apparently, there are references to OpenStreetMap in the firmware (Meta joined as a Strategic Member last year) and it seems unlikely that Zuckerberg would cede that territory to Apple or Google.

Mapping represents an important battleground for spatial computing generally and for spatial gaming particularly. My co-author Afo and I touched on this in The Road Ahead for Spatial Gaming where we wrote:

For spatial gaming experiences to persist across sessions and users in public spaces across the world, one or more layers of ubiquitous virtual positioning systems will need to emerge.

We’ll have to wait and see how Meta approaches this, but it is far from the only company tackling this challenge. In June, the newly spun-out Niantic Spatial announced a “strategic multi-year partnership” with Snap “to jointly build a next generation AI map that will serve as the foundation for AR glasses and AI agents to understand, navigate and interact with the real world.”3

The “AI” is definitely doing a lot of heavy lifting in this quote — presumably on behalf of Snap’s investor relations team — but nevertheless, Snap is a noteworthy competitor to Meta’s spatial computing ambitions (and is already stepping on it’s social media turf, to boot). The announcement quotes 400,000 AR developers in the Snap ecosystem serving 900 million Snapchatters, so it’s definitely not nothing.

If the look of the glasses is any indication, though, Snap has bigger problems than competition with Meta to worry about. Say what you will about the Meta Ray-Bans, but they are at least passable as a normal pair of thick glasses. The Snap Spectacles look like a cross between swim goggles and Geordi La Forge’s VISOR.

Either way, there is sure to be more to come on maps and smart glasses in the near future.

In a story that should come as little surprise to readers of Always Scheming, generative AI usage in game development is growing rapidly.

Ichiro Lambe of Totally Human Media (operators of game discovery platform We Love Every Game) recently conducted an analysis on the disclosed usage of AI among Steam titles (via the Steam API) and found meaningful growth over the last year.

In his initial analysis back in April 2024, Lambe observed about 1,000 titles disclosing usage of generative AI, or roughly 1.1% of the overall Steam library. This latest follow up saw the number rise to 7,818 titles, or about 7% of the total Steam library. When looking at just this year, Lambe observed that nearly one in five games released in 2025 have disclosed AI usage.

According to Lambe, common use cases for AI include generation of visual assets, audio assets, text, and narrative, as well as creation of marketing and promotional materials. While there was no explicit mention of vibe coding, generation of code and game logic was cited as another frequently disclosed use case.

One in five feels like a large share, given how early we are in the evolution of these tools. However, consider that there may be many games that used generative AI in one form or another that didn’t actually disclose it. How much higher might the number be if we were able to capture those titles, too?

Beyond simply missing the step or forgetting to mention it in publishing materials, there are reasonable arguments for avoiding disclosure given the vociferous anti-AI backlash among many on social media. Just ask the folks at 11 Bit Studios.

Steam isn’t going anywhere, nor is generative AI, so expect this number to keep rising. However, I’m particularly interested in where the vibe coded games land. Steam doesn’t feel like the right platform for them: they’re too ephemeral, too meme-based (see: Outpunch, Coldplay Canoodlers, Gorilla Warfare, etc.), too janky.

Put differently, AI usage among Steam games only captures part of the picture. There’s a whole new class of interactive entertainment that is almost entirely orthogonal to our traditional notions of how games get made and consumed. Whether those games can capture value remains an open question, but even Gaben himself is thinking about it:

“… I think we'll be in this funny situation where people who don't know how to program who use AI to scaffold their programming abilities will become more effective developers of value than people who've been programming, y'know, for a decade."

As part of the monthly Field Notes cadence, I want to engage with you all a bit more (yes, you!) and get a better pulse on what’s showing up on your radar.

Every month, we’ll have a new poll to chime in on. Over time, we should be able to surface some interesting data that I’ll then share back with you all at a regular cadence.

That’s it for this month!

Thanks for tuning in. Back with more Field Notes in August.

Until then.