Tech Landscape #433
Apple tries AI again, LLMs become a political football, and lots of little social updates.
Hello!
Photography is the easiest way to create content today. As long as you have a camera, you can just point and click, and you have an image. But not every photograph is a good photograph. Your smartphone camera will run all sorts of algorithms to make your photograph look better, but there’s no guarantee of a good photograph. You can buy yourself the best camera and accessories in the world, and it’s still no guarantee of a good photograph. What makes a good photograph is that indefinable ‘photographer’s eye’; making choices informed by discernment and taste, knowing what to capture and how and when to capture it.
Yes, this is an analogy for creating content with AI.
Anyway, let’s get on with it. Hope you’re well!
Apple WWDC 2026
At its annual developer conference Apple announced several new things including lots of performance and user experience improvements coming to all of its software platforms and new child safety features such as parent-controlled Child Accounts with contact and content filters.
But the bulk of the announcement covered new Apple Intelligence features, including improved photo editing and an upgraded Image Playground app with photorealistic generation, and vibe-coded Shortcuts to automate actions across apps, all with Apple’s enhanced privacy layer.
At the heart of it is the new Siri AI, “a profoundly more capable and personal assistant” that features conversational context understanding, onscreen awareness, and cross-app action capabilities, and works on iPhone, MacOS, VisionOS, and more.
For users of Android phones (especially Pixel) this won’t be anything new; it’s largely Apple playing catchup. But for many people this will be their introduction to the capabilities and possibilities of having AI deeply integrated into a device. It has some nice Apple-y touches, including running many common tasks using on-device models (as long as you have a very recent device) and handing off to cloud servers for more complex tasks.
Siri AI and the new Apple Intelligence features are available in beta to test now, and will become generally available later this year — unless you live in the EU, where Siri AI will be ‘delayed’ due to Apple’s non-compliance with the DMA. Apple thinks its too risky to allow other assistants full access to a user’s data, which is sort of classic Apple because yes, that’s understandable, but it also privileges Apple’s own apps again and takes away the choice of users who might actually want that.
I haven’t tried any of the new features myself, so if you want more informed insights you should watch Marques Brownlee’s first impressions.
Synthetic Audio-Visual
Luma released Ray3.2, an upgraded video generation model featuring multi-keyframe control, 1080p video-to-video editing, and improved motion transfer. lumalabs.ai
I made a few quick tests ⬇️; the detail and camera motion are excellent, but its world understanding is off. It seems to be a fast and decent option, even if not consistently up with the best.
Riverflow released v2.5 of its reasoning image model, adding customised ‘thinking’ levels, better font control, and optional plain or transparent backgrounds. riverflow.ai
Creative Tools
MiniMax launched Hub, a desktop application for creative workflows using teams of agents and local files. hub.minimax.io
LetzAI introduced Canvas, an infinite board interface that lets users work in a unified creative workspace. letz.ai
ElevenLabs’ ElevenCreative introduced Avatars, a new feature that combines its speech models with lip-syncing technology in a single interface to let users generate talking-head videos. elevenlabs.io
Magnific added Auto Layers, a tool to convert static images into layers for easier editing. x.com/magnific
This is a very useful feature, the type of thing that can get more digital artworkers and designers on board.AI music app Mureka launched a new version. instagram.com/mureka.ai
I know that’s a skimpy description but some companies are just awful at promoting their changes. There isn‘t even anything in the What’s New section of their app, so I don’t know what I can tell you.
Social
Two small but useful updates in Instagram: Your Algorithm choices now work on the Feed, and manual reordering of the profile Grid.
Edits’ latest update includes Reels Search, project versioning, and a tab to enable Beta features. threads.com/@creators
Meta held a creator event this week where it showed off more features coming to Edits, including a desktop app and an AI assistant.Bluesky added Group Chat, allowing up to 50 users in public or private groups. bsky.app/@bsky
YouTube is expanding its in-app DMs to more countries. blog.youtube
Discord launched You Bar, a new mobile navigation that simplifies the app navigation, highlights user profiles with larger avatars, and aligns the mobile interface closer to desktop. discord.com
Pinterest launched Amazon Storefront linking, letting eligible creators connect their Storefront account to their profile and automatically appliying affiliate links when they tag Amazon products. newsroom.pinterest.com
Snapchat launched a friends-only profile for users aged 13 to 15, restricting their Stories and Spotlight videos to mutually accepted friends and removing public distribution for this age group. newsroom.snap.com
All the big social platforms (bar the obvious) are adding more child safety features in the face of increasing regulation. Meta thinks this should be done at the operating system level.
Assistants & Search
Claude Fable 5 and Claude Mythos 5
Anthropic launched two new models with advanced reasoning, coding, and vision skills, saying their “capabilities exceed those of any model we’ve ever made generally available”. Fable 5 is built for general use with strict safety guardrails, and Mythos 5 has reduced guardrails and is available only to trusted cyber-defenders. anthropic.com
Shortly after it launched, the US government put out a directive ordering Anthropic to limit access by any non-US nationals — including Anthropic’s own employees. Unable to implement this, the company pulled access from everyone. It’s not clear why this directive was issued, although the government’s ‘AI Czar’ David Sacks said “a trusted partner” (reported to be Amazon) had managed to jailbreak the safeguards, potentially exposing advanced capabilities to anyone. Anthropic denies this.
Whatever the truth is, two things are clear: that advanced models have become a matter of national security and/or protectionism; and that companies who persistently tell people that their models are too dangerous to release, but then release them anyway, might get caught out if someone decides to take them at their word.
Anyway, I don’t have enough spare cash to have tried this for myself yet, but people had been vibe-coding amazing things with it before the ban, like this MMORPG World of ClaudeCraft. To get a better sense of it I recommend you read Ethan Mollick’s thoughts on what it feels like to work with Mythos.
Google expanded Gemini in Chrome to desktop and iOS users in Latin America, Africa, and the Middle East. blog.google
… but not in the EU or UK yet.