Retter Reboot #17

The Retter Reboot : Your weekly dose of ai & tech news 

Listen Our Podcast 

OpenAI Ships GPT-5.5 — One Step Closer to the Long-Promised AI "Superapp"

OpenAI released GPT-5.5 on April 23, calling it their "smartest and most intuitive to use model" yet. Brockman explicitly tied the release to the company's superapp ambition: ChatGPT, Codex, and an AI browser merged into a single enterprise tool. GPT-5.5 benchmarks above both Claude Opus 4.6 and Gemini 3.1 Pro (per OpenAI). Faster, sharper, fewer tokens than 5.4 — but 2x the API price. Available now on Plus, Pro, Business, and Enterprise tiers, with GPT-5.5 Pro on Pro/Business/Enterprise. Codex gets a massive upgrade: 400K context, browser use, and 5.5 Fast mode.

Source: TechCrunch

Anthropic Launches Claude Design — And Figma's Stock Drops 7% the Same Day

On April 17 Anthropic shipped Claude Design, a text-to-prototype tool that builds design systems, interactive websites, slide decks, and one-pagers from prompts — powered by the new Opus 4.7. It reads your codebase, extracts a design system, and hands off directly to Claude Code for production. Exports to Canva, PDF, PPTX, and HTML. Figma dropped ~7% on the news; CPO Mike Krieger had resigned from Figma's board three days earlier. Available in research preview on Pro, Max, Team, and Enterprise. Anthropic's move from foundation model provider to full-stack product company is no longer subtle.

Source: VentureBeat

Claude Now Connects to Spotify, Uber Eats, TurboTax, AllTrails & More — Personal Life Joins the Work Stack

Anthropic expanded Claude's connector directory beyond work tools with a consumer-focused wave: AllTrails, Audible, Booking.com, Instacart, TurboTax, Resy, Spotify, StubHub, TaskRabbit, TripAdvisor, Uber, Uber Eats, Viator, and more. The directory now tops 200 connectors since launching in July 2025. Claude dynamically surfaces the right app in conversation, asks confirmation before booking or purchasing, and emphasizes no ads, no sponsored placements, and no training on connector data. Available on all Claude plans, with mobile in beta.

Source: The Verge

Amazon Pours Another $25B Into Anthropic — Mirroring Its $50B OpenAI Bet

Amazon is investing up to $25B more in Anthropic ($5B now, $20B tied to milestones), on top of the $8B it has already put in since 2023. Anthropic commits to spending $100B+ on AWS over 10 years and gets up to 5 GW of Trainium capacity. The full Claude Platform will be available directly inside AWS — a deeper integration than Bedrock. This mirrors Amazon's $50B/$100B deal with OpenAI two months ago: Andy Jassy is placing parallel mega-bets on both frontier labs. Direct rebuttal to OpenAI's claim last week that Anthropic "made a strategic misstep not to acquire enough compute."

Source: CNBC

Google Cloud Next: Gemini Enterprise Agent Platform Launches With Claude Opus, Sonnet & Haiku Support

Google unveiled the Gemini Enterprise Agent Platform at Cloud Next on April 22 — the evolution of Vertex AI and the company's full-stack bet on the agentic enterprise. The interesting choice: Google is offering Anthropic's Claude Opus, Sonnet, and Haiku (including the new 4.7) as first-party model options alongside Gemini and Nano Banana 2. Ships with a no-code Agent Designer for business users, a graph-based Agent Development Kit for devs, plus Agent Runtime, Memory Bank, and a 200+ model garden. Google is betting enterprise customers want the platform, not the pieces.

Source: TechCrunch

OpenAI's ChatGPT Images 2.0 Tackles Charts, Diagrams & Multilingual Text

OpenAI rolled out ChatGPT Images 2.0 (gpt-image-2) on April 21 — a major upgrade targeting the productivity crowd. The pitch: accurate complex charts, scientific diagrams, structured layouts, floor plans, image grids, manga sequences with character continuity, and multilingual text that reads correctly in context. Supports up to 2000px-wide output and 8+ images in one prompt. Includes a "Thinking" mode where the model reasons, plans, and does web research before rendering. A clear signal: OpenAI is done with consumer creative moonshots (Sora is shuttered) and chasing professional workflows.

Source: Bloomberg

Vercel Breach — Supply-Chain Attack via Context AI OAuth Exposes Customer Data

Vercel confirmed a security incident: a "ShinyHunters"-branded threat actor accessed internal systems, allegedly selling source code, API keys, GitHub tokens, and NPM tokens for $2M on BreachForums. Root cause: a Vercel employee connected the Context AI Office Suite (compromised via Lumma Stealer in February) to their enterprise Google Workspace with "Allow All" OAuth permissions. From there, the attacker pivoted into Vercel. Next.js and the public package supply chain are confirmed clean. The lesson is bigger than Vercel: OAuth apps are the new supply-chain attack surface — and "Allow All" is the new "rm -rf".

Source: BleepingComputer

Musk's xAI Chases Anthropic With Three-Way Alliance Talks — Mistral + Cursor

Business Insider reports xAI has held three-way partnership talks with Mistral and Cursor, on top of SpaceX's $60B option to buy Cursor disclosed Tuesday. Mistral cofounder Devendra Chaplot already leads xAI pretraining; Cursor is training on xAI's Colossus cluster in Memphis. Musk has reportedly told engineers he's worried about Anthropic's lead in coding and agents; xAI president Michael Nicolls said the company is "clearly behind" rivals. Context: Anthropic blocked xAI's access to Claude through Cursor in January. The Musk stack is consolidating — not just a model lab, but compute, editor, and European frontier research under one umbrella.

Source: Implicator.ai