Tech Landscape #434
New XR glasses at AWE, the State of Unreal, and the UK’s teen social media ban.
Hello!
I’m trying to get a better grip of agentic workflows and MCP, so I paid for a Manus subscription and set myself a starter task: to create this week’s cover illustration. These were the steps I settled on:
Check the Airbase table [via MCP] where I store each week’s stories, and get all of the written summaries for each story.
From those, synthesise a cover concept, research [via the Web] current design and illustration trends, and turn that into an image prompt.
Send the prompt to Magnific [via MCP], choosing the image model best suited to the task
After testing and revising it a few times, I turned it into a repeatable Skill. And here’s the result:
Clearly it isn’t great; it’s a bit overly-literal in some places, overly-metaphorical in others, needs more of a high concept. But I’m learning a lot about how to prompt multi-step workflows, which is useful.
If you’d like to contribute to the cost of Airbase, Manus, Magnific, and the other tools I use and test, consider becoming a paid subscriber.
Right, let’s get on with it. Hope you’re well!
AWE 2026
The annual Augmented World Expo took place this week, where a lot of brands you’ve never heard of, and a few you have, showcased new products and features. These were the biggest announcements:
Snap introduced SPECS
The company’s first standalone (no puck, no tether) augmented reality glasses feature a 51° field of view that can show a near 24” or far 110” virtual screen, with hand-tracking, an AI assistant, Lenses, and “up to four hours of mixed-use battery life”. They’re available for pre-order at $2,195, coming to the UK, US, and France this Autumn.
newsroom.snap.com/introducing-specs-augmented-reality-glasses
The technology sounds impressive, but there’s no getting around the fact that these look pretty bulky, and that price… I appreciate the fact that Snap is pushing the market forward, but I’m not really sure who these are for. £2000 to watch videos and play with Lenses? I hope they’re in it for the long haul.
XREAL revealed AURA
The ‘spatial computing’ glasses are powered by Android XR and the Snapdragon Reality Elite Platform (⬇️), with 70° FOV, hand-tracking, and a tethered puck for compute and battery.
I was impressed by the XREAL One glasses I tried; the image was sharp, the audio clear, and they were comfortable. AURA seems more of a generalists device than the Snap SPECS, and comes with the advantages of access to Android apps and a price expected to be around $1500. That’s still a little spicy, but more attainable than Snap’s glasses. Pre-orders are open now, planned to ship in Autumn to the UK, US, Canada, Japan, and South Korea. Read more on the AURA product page, or watch Marques Brownlee’s hands-on.
Qualcomm announced the Snapdragon Reality Elite Platform, a next-generation chipset built for “AI‑enhanced XR” on spatial computing devices with big (potential) CPU, GPU, and NPU improvements over its previous chipset. qualcomm.com
Social
Facebook added AI Mode, an answer engine for searches grounded in public discussions across Groups and Reels. It also improved camera roll sharing suggestions and added new AI photo presets. about.fb.com
A search engine powered by Facebook posts? 😬
Threads celebrated reaching 500m MAUs with Your Algo, which is a new way for users to control the topics they see in their feed, and graduating Communities out of Beta with a few new features. about.fb.com
Instagram now lets each item in a carousel have its own caption. instagram.com
Mastodon 4.6 introduced Collections for curating and sharing lists of other users — kind of like ‘starter packs’ of who to follow. It also lets select users post to email, not quite a newsletter tool but a step towards it. blog.joinmastodon.org
Telegram’s latest updates include letting bots reply with rich text, AI ‘guardians’ to manage entry to group chats, and Apple Watch and Wear OS apps. telegram.org/blog
Safety
The UK plans to ban social media platforms for under-16s. The ban will include “user-to-user platforms, whose purpose is to enable social interaction and which allow users to post material, alongside algorithms — meaning, at least, Snapchat, TikTok, YouTube, Instagram, Facebook, and X. The legislation is expected before Christmas, with the ban coming into force in Spring 2027. gov.uk
Will this be effective? Opinions are very divided, and not along the “anti-tech / pro-tech” lines you might expect. Will young people find it a benefit? Will they find ways around it? Will it push them to less regulated platforms such as Discord and Telegram instead? I don’t know. Nobody knows. And that’s the problem; we had the opportunity to wait until proper data came from Australia, and we didn’t take it.
Meta expanded Teen Accounts safety features, rolling out the 13+ content settings globally and adding AI-powered age detection and parental alerts for self-harm searches. about.fb.com
Google is expanding Pixel’s parental controls to all devices that update to Android 17. blog.google
Ads & Commerce
Meta announced new product discovery features for businesses including Live Shopping Tools on Facebook, the expansion of Live Video Ads to Instagram, and Virtual Cards for more private purchases. facebook.com
Pinterest introduced new AI-powered tools including Business Assistant, an MCP server so you can use it with other assistants, and new Pinterest Performance+ creative capabilities. It also announced Ask Pinterest, an experimental app for AI-powered shopping experiences. newsroom.pinterest.com
Snapchat introduced new AI-powered features for businesses including the Snap Smart Assistant for campaign setup, an MCP server, generative creative ad tools, and Snap Creator Network that uses an AI system for creator matchmaking. forbusiness.snapchat.com/blog
Gaming & Metaverse-ish
State of Unreal
Epic Games held its annual developer event this week. The headline announcement was the release of the latest version of its creative engine, Unreal Engine 5.8, with all kinds of improvements for gaming and production, including markerless motion capture (based on the acquisition of Meshcapade [TL 418]), AI post-rendering, and support for MCP that enables developers to integrate agentic tools (e.g. Claude Code) into their workflows. Plus there was news that The Simpsons is the next licensed IP that creators can work with, and that over $1bn has been paid out to Fortnite developers since UEFN was released.
unrealengine.com/news/state-of-unreal-2026-top-news-from-the-show
It was also revealed that the next version, UE 6, will unify UE and UEFN (the Fortnite editor), with the aim of making “content, code, and economies… become portable and interoperable across games, ecosystems, and engines through open standards” — the metaverse dream is still alive!
Roblox launched its age-based Kids and Select accounts globally, automatically placing users into age-appropriate tiers with tailored game catalogues, chat restrictions, and parental controls. about.roblox.com
Electronic Arts launched EA Advertising, a platform that enables brands to integrate directly into gameplay and live experiences across its global portfolio. news.ea.com
Synthetic Content
Bytedance launched Seedance 2.0 Mini, a cheaper, faster version of its flagship video model. instagram.com/dreamina_ai
It’s available exclusively for a limited time in ByteDance’s Dreamina creative platform before rolling out more widely.Vidu introduced Q3 Mix, an upgraded AI video generation tool with enhanced lens control capabilities including intelligent shot switching, full camera movement, and advanced shot type understanding. linkedin.com
Indian startup Avataar launched Varya, the first video model trained with extra context for the local market. techcrunch.com
It’s based on Wan 2.2 so isn’t the most modern, but also isn’t a compute resource hog.Mozart added Personas, to create a custom voice model that can be used for vocal consistency. instagram.com/mozartaiofficial
Alibaba ATH released HappyOyster 1.0, a world model with both directable video and adventure (’video game’) modes. x.com/HappyOysterAI
Here ⬇️ is a recording of a test I made; it’s three minutes long so feel free to skip forwards a few times. It starts with the man walking, then I ‘directed’ (added events to) it through prompting. It’s far from perfect, and loses coherence the longer it runs, but it’s still pretty amazing that this is becoming possible now. (Which is why I find it really weird that a technological breakthrough like this is only promoted through a single post on X.)
Creative Tools
Adobe’s Firefly Assistant gained new capabilities, including skills for brand kit and product video creation, workflow preferences, and sharing with collaborators. It also teased a forthcoming redesign / upgrade with reusable elements and project organisation, currently in closed Beta. blog.adobe.com
Adobe’s AI Assistant is rolling out to major Creative Cloud applications. blog.adobe.com
Higgsfield launched a Photoshop plugin with real-time generation, automated layer splitting, and image generation. threads.com/@higgsfield.ai
Framer’s 3.0 release introduced native AI Agents, an isolated Branching system for teams, and an all-new Community ecosystem with an open Marketplace. framer.com
Anthropic updated Claude Design with better support for custom design systems, integration with Claude Code, a new canvas editor, and more connectors to third-party platforms. claude.com
Luma Agent added Skills, repeatable creative workflows you can build once and run again. instagram.com/dreamlabla
Creative applications add agents; agentic platforms add creative applications. Somewhere in the middle of that is the new creative interface.
Epic Games released a video showing how generative AI is integrated into its art concepting process for Fortnite artists:
I find it really interesting to see AI adopted into professional workflows. But, as you might expect, the comments are ridiculous. Yes, it makes occasional mistakes; but undoing those mistakes is still faster than doing everything from scratch. I don’t know how people can watch this video, with the amount of human input and guidance it shows, and then still dismiss it as slop. “If there’s any AI, it’s slop” isn’t criticism, it’s dogmatism shored up by bad-faith arguments.
Assistants & Search
Google Search rolled out Information Agents, which can passively monitor topics, tasks or projects and send updates when they change. threads.com/@google
Available to AI Ultra subscribers worldwide.Microsoft’s Copilot Cowork is now generally available. microsoft.com
The agentic system designed for complex, long-running, multi-tool tasks is priced on a per-usage basis, which is going to rack up some bills.Perplexity announced Brain, a self-improving memory system that reviews previous actions and learns how to do them better. perplexity.ai
This seems similar to Manus’ self-updating projects and Claude’s ‘dreaming’ [TL 429] and ChatGPT’s Dreaming [TL 432] — clearly an idea whose time has come.
Everything Else
Google released Android 17, bringing enhanced multitasking, on-camera reactions in screen recording, an optimised gaming mode for foldables, and new security and privacy features. blog.google
New Android version releases are fairly meh nowadays, as Google releases year-round updates. But if you’re a dev who wants to make the most of the new capabilities you should check out the developer blog post.