Multimodal

Meta launched Muse Image on July 7, built by Meta Superintelligence Labs and free in Meta AI app, Instagram Stories, and WhatsApp. It does standard AI image generation with preset prompts. The flashpoint: you can @ any public Instagram user and remix their photo into a new AI image. The post doesn't say whether users can opt out or what training data was used. I'd hold off on trusting the privacy story.

Meta's Superintelligence Labs released Muse Image and Muse Video. Muse Image handles precise instruction-following, editing, and multi-reference composition using Instagram social context, plus agentic tool use. Muse Video shares the same pretrained base, outputs video with native audio. Available now in select countries via Meta AI app, web, Instagram Stories, and WhatsApp. The post doesn't disclose model size, latency, or which countries.

ByteDance's Seed team released Seedream 5.0 Pro, a multimodal image model built for design tasks. It turns data, timelines, and charts into usable infographics with accurate dense text. Editing supports point-and-click, lasso, sketch-to-render, layer separation, and multi-image compositing—allowing pixel-level changes without regenerating the whole image. Portrait and material quality are improved, handling skin texture, glass reflections, and panning motion blur. The model natively supports input and text rendering in over ten languages, including right-to-left Arabic. It is live on Volcano Engine's experience center and will roll out to Doubao and Jimeng.

Anthropic re-released Fable 5 for paid users with stronger guardrails, available in subscriptions only through July 7 and capped at 50% of usage limits. Scale's benchmark shows it completes 16% of remote work tasks, double Opus 4.8. Claude Sonnet 5 also launched—benchmarked close to Opus 4.8 on agent tasks, cheaper per token but roughly the same cost per task in practice; the author finds it expensive and slow. Google dropped two new models: Nano Banana 2 Lite for fast, cheap images and Omni Flash for video generation and editing. Bridgewater and Thinking Machines trained a financial triage model hitting 84.7% accuracy at 13.8x lower cost than the best frontier model tested.

Day 3 of AIEWF focused on autoresearch. Introspection's Roland Gavrilescu described it as an outer loop where agents maintain the system itself. Anthropic's Thariq Shihipar echoed continuous discovery in his Claude Code keynote, saying models are 'grown, not developed.' Former Google engineering lead Addy Osmani pushed back hard: the outer loop must stay human—inner loop is capability, outer loop is agency. Notion's Geoffrey Litt and Impeccable's Paul Bakaus both argued humans need to understand the code and steer the final 20%. Bakaus stated flatly there will 'never be auto.' Google's Nicole Brichtova added that cultivated expertise sees what average preference misses.

Blogger Xiaohu released an open-source tool called 'Xiaohu IP Studio' that auto-generates illustrations for articles. It ships with 31 original characters—15 hand-drawn line-art figures and 16 pun-based meme images. The agent reads the article, decides on an illustration type (mood image, diagram, or four-panel comic), generates the image, and self-checks with rework if needed. The default style is hand-drawn line art with light color; five alternative skins are available, including 3D blind-box and black-and-white line art. Setup requires only Python 3, works with Claude Code or Codex, and needs an OpenAI-compatible image API key (defaults to GPT-image-2). You can also output prompts only and generate images manually.

lcamtuf bought a mid-2026 AI-generated children's encyclopedia that was an Amazon #1 bestseller. The illustrations are full of body horror: extra-limbed cats, fused animal-tree masses, and distorted human reflections. He argues these books sell because buyers don't read them, covers fool gift-givers, and there's no IP risk. The post doesn't assess factual errors in the text, but the images alone are alarming.

Google added a native computer-use tool to Gemini 3.5 Flash, letting the model take screenshots, click buttons, and fill forms to operate web and desktop UIs. It joins Anthropic and OpenAI in baking screen control directly into a model. The post claims 3.5 Flash beats Claude Sonnet 4 and GPT-5 on the WebVoyager benchmark, but Google didn't release full eval details or reproduction steps—hold for third-party tests. Available now via Gemini API and Google AI Studio; pricing and rate limits aren't disclosed in the post.

Some users already see Bidi 1 in ChatGPT's web and app model selector, sitting alongside Standard and Advanced Voice. Selecting it turns the voice bubble yellow. The key change is full-duplex: the model can keep listening while it speaks and respond immediately to interruptions. In a demo, a user asked it to count from 1 to 10, interrupted mid-way and told it to count backwards—it complied instantly. OpenAI hasn't announced a launch date; one outlet speculates a wider rollout this week. The post doesn't disclose pricing, regional availability, or a firm timeline.

An FT comment argues that OpenAI's image model, trained on Shutterstock, accidentally generated photos with Getty Images watermarks. The blunder became Getty's best ad: high-quality licensed content can't be replaced by synthetic data. The more AI firms rely on cheap stock libraries, the stronger Getty's scarcity premium looks.

ByteDance Seed team released the Seed2.1 model series, now live on Doubao and TRAE. The update focuses on getting real work done rather than static benchmarks. For general agent tasks, Seed2.1 Pro ranks in the top tier on Agents' Last Exam, achieves top score on MobileWorld for phone GUI tasks, and cuts average steps for cross-tool tasks by 16%. In coding, Seed2.1 Pro wins 59.1% of blind developer evaluations against Claude Opus 4.6 and ranks 8th on the Code Arena frontend leaderboard. Multimodal understanding hits SOTA on CharXiv-RQ, TVBench, and others. The team also uses Seed2.1 agents internally for data synthesis and training optimization. The post does not disclose parameter count, pricing, or max context window.

Runway plugged its flagship video editor Aleph 2.0 into Figma's creative canvas tool Weave. You extract a frame, restyle it, attach a timestamp, and Aleph 2.0 propagates that edit across every frame where the subject appears—everything else stays untouched. It handles 30-second 1080p clips and multi-shot sequences without frame-by-frame work. Inside Weave it's just another node you chain with image gen and compositing tools. Swap a product, change a background, or restyle a whole scene, with single-frame previews before you generate.

San Francisco agency Qontour copied the full text of John Koenig's 'The Dictionary of Obscure Sorrows'—all 311 neologisms and the foreword—onto a site they built, added DALL·E 2 illustrations and a GPT-4 word generator. Koenig had no involvement. The domain differs from the original by one 'the,' and the footer admits they don't own the rights.

Midjourney unveiled a full-body ultrasound CT scanner in San Francisco. A person stands in a water ring while 40 Butterfly ultrasound-on-chip modules emit sound waves from all directions; 21 servers reconstruct 3D cross-sections with 2 PFLOPS of compute. The scanner cannot image brain, lungs, or bowel—it targets body composition analysis under FDA Class II. The first location is a spa in Union Square, not a hospital sale. The real story is the funding: Midjourney has zero outside investors. Subscription revenue from millions of image-generation users covers a $15M Butterfly upfront payment, $10M annual licensing, a 9-person hardware team, and the Union Square lease. Founder David Holz's VC aversion traces back to his previous startup Leap Motion, which raised over $100M and sold for roughly $30M. Midjourney's actual revenue is undisclosed; third-party estimates range from $200M to $500M ARR. Holz mentioned a speculative $20B capex for 50,000 units—even optimistic free cash flow would need 80 years, so outside capital likely becomes necessary at scale. Only 12 people have been scanned, no clinical data is published, and Butterfly carries $879M in cumulative losses, creating a single-point supply-chain risk.

Adobe is putting its 'creative agent' into Photoshop, Premiere, Illustrator, InDesign, and Frame.io in public beta. Users describe the end result and the agent handles multi-step grunt work: rough cuts, clip sorting, background swaps, batch resizing, and generating 50 file versions from a spreadsheet. Firefly gets new solo-creator tools like a brand kit and product-photo-to-video. Adobe tools are already usable inside ChatGPT, Claude, and Microsoft 365 Copilot, with Google Gemini and Slack integrations coming. The After Effects assistant remains in private beta; the post doesn't give a public release date.

Mindgard researcher Jim Nightingale found that a viral prompt bypasses ChatGPT's image generation filters, causing it to spontaneously produce sexual violence and snuff imagery. The prompt simply asks ChatGPT to 'restore the attached photo' without specifying content, yet the model generates extremely graphic images involving death and sexual assault. Nightingale had previously reported nude image generation to OpenAI, which claimed the issue was resolved. The post does not disclose OpenAI's response timeline or specific remediation plans for this new finding.

Midjourney rolled out Draft mode for V8.1: click the lightning button to generate 24 low-res previews at half the fast-hour cost of a standard job. Pick the ones you like and hit Vary to render them at full quality. A new --preview flag also lets you test early model versions, though outputs may be rough and jobs aren't guaranteed to stay consistent—differences are most noticeable with personalization and moodboards. The post doesn't disclose Draft mode's exact resolution or which model --preview points to.

Bloomberg published the full letter Commerce Secretary Howard Lutnick sent to Anthropic. The letter demands an explanation for why Mythos could generate deepfake images of Trump and Musk, and questions the content moderation system. Anthropic then voluntarily disabled Mythos's image generation. The article doesn't say whether the shutdown is temporary or permanent, and gives no timeline for restoration.

At a closed-door session after WWDC, Apple revealed they had a working prototype last year—an improved old Siri with tool calling. The team decided it fell short of the product vision, so they scrapped it and rebuilt the entire architecture from scratch on a new large model. The new Siri has a standalone app, native multimodal support, privacy built into the foundation, and runs the same system across iPhone, iPad, Mac, Apple Watch, Vision Pro, CarPlay, and AirPods. Mike Rockwell took over Siri management last year.

Bloomberg's Mark Gurman tested the new Siri in iOS 27 and macOS 27. It can understand on-screen context and perform cross-app tasks—like finding a photo, editing it, and sending it via Messages—with a single voice command. Complex tasks still take 11+ seconds and occasionally miss steps. Gurman calls it 'just good enough': a big leap from the old Siri but still trailing Google Astra. The post also mentions a foldable iPhone and touchscreen MacBook in development, with no release dates disclosed.