Anthropic Pushes Beyond Chatbots

Plus, ✨How to generate and edit images for free with HiDream-O1, Cursor trained Composer using dramatically larger synthetic coding task datasets, and more!

In partnership with


Welcome back to The Shift. Let’s get straight to what matters in AI today…

Today we have:

🧪 Anthropic and SandboxAQ expand beyond chatbots

✨How to Generate and Edit Images for Free With HiDream-O1

🧠 Cursor upgrades Composer for longer AI coding workflows

🔨Tools and Shifts you cannot miss

🧪 Anthropic and SandboxAQ expand beyond chatbots

Anthropic and SandboxAQ announced major updates focused on scientific AI and developer infrastructure, with SandboxAQ bringing advanced research models into Claude while Anthropic tightened control over AI agent integrations through its Stainless acquisition.

The Shift:

Scientific models become conversational - SandboxAQ integrated its quantitative scientific AI models directly into Claude, allowing researchers to run advanced drug discovery and materials science workflows through simple natural language interactions.

Researchers no longer need infrastructure - SandboxAQ said users previously needed specialized computing environments to access these systems, but Claude now removes much of the technical complexity around scientific model deployment and experimentation.

Developer infrastructure moves in-house - Anthropic acquired Stainless, the SDK automation startup used by OpenAI, Google, Cloudflare, and other rivals, giving Anthropic deeper control over critical AI developer infrastructure tooling.

Anthropic closes ecosystem loop - Anthropic said Stainless will wind down hosted products for outside companies, meaning infrastructure previously shared across the AI industry will now become tightly integrated into Anthropic’s own ecosystem.

AI companies are increasingly competing far beyond chatbot quality alone. The race is now shifting toward owning developer infrastructure, scientific workflows, enterprise tooling, and the systems researchers depend on behind the scenes.

Together with Wispr Flow

Talk to your AI tools the way you'd talk to a colleague.

You don't send a colleague a three-word brief. You explain the context, the constraints, what you've already tried. But typing all that into ChatGPT takes forever — so you don't.

Wispr Flow lets you speak your prompts instead. Talk through your thinking naturally and get clean, paste-ready text. No filler words. No cleanup. Just detailed prompts that actually get you useful answers on the first try.

Millions of users worldwide. Works system-wide on Mac, Windows, and iPhone.

How to Generate and Edit Images for Free With HiDream-O1

An open source image model with word accuracy on par with GPT-Image-2 and Nano Banana. 8B parameters. MIT License. Here's how to use it.

Step 1: Try it instantly on Hugging Face - No setup needed. Head to the hosted demo and start generating immediately, no GPU, no install, no account required.

Step 2: Generate from text - Describe your image in plain language. HiDream-O1 handles photorealism, long text rendering, diagrams, ads, and posters with strong prompt adherence out of the box.

Step 3: Edit existing images - Upload an image and describe what to change. Add, replace, or remove elements directly, no manual masking or inpainting tools needed.

Step 4: Generate storyboards - Feed in a sequence of prompts and let HiDream-O1 build consistent scene-by-scene visuals automatically.

Step 5: Run it locally - Download the model from Hugging Face and self-host for unlimited generation with zero API costs. MIT license, use it commercially, modify it, build on it.

8B parameters. On par with closed-source models that cost orders of magnitude more to run. You can try the demo here. or download the model here.

🧠 Cursor upgrades Composer for longer AI coding workflows

Cursor released Composer 2.5, a major update to its coding model focused on improving long-running task execution, instruction following, communication quality, and reinforcement learning behavior across complex software workflows.

The Shift:

Cursor improves coding reliability - Cursor said Composer 2.5 performs better on sustained coding sessions, handles complex instructions more reliably, and improves collaboration behavior during longer multi-step software development workflows.

Composer undercuts frontier rivals - Cursor said an average CursorBench task costs under $1 using Composer 2.5, compared to reportedly reaching around $11 for Opus 4.7 and GPT-5.5 at similar performance levels.

Training complexity increased sharply - Composer 2.5 was trained using 25x more synthetic coding tasks than Composer 2, including feature deletion challenges built from real-world codebases and functional software testing systems.

Larger model training begins - Cursor revealed Composer 2.5 was partially trained using Colossus 2 infrastructure, while also training a significantly larger SpaceXAI model using roughly ten times more compute resources.

AI coding models are becoming increasingly autonomous during long software workflows instead of only responding to isolated prompts. The bigger challenge now is not just improving capability, but controlling unexpected behaviors emerging during advanced reinforcement learning training.

Together with SureThing

One agent, one brain, zero manual work.

Most AI tools forget you the moment the chat ends. SureThing doesn’t.

SureThing is an autonomous agent that can draft in your voice, triage what matters, follow up on things you forgot, and report back with what happened next.

Day 1, you onboard it.
Day 30, it knows your clients and patterns.
Day 90, it catches things you missed.


🔨AI Tools for the Shift

🚀 UGO – Automatically creates social posts, blogs, SEO pages, and ad campaigns using content tailored to your actual brand voice.

🎵 MusVideo – Turns uploaded audio into cinematic AI-generated music videos without requiring editing or directing skills.

🎬 Flyne AI – Converts photos into engaging AI-generated videos for content creation and storytelling.

💇 Silky AI – Lets you instantly try different hairstyles with AI before making a real-life change.

📹 Vidlogi – Generates fully produced SEO-focused videos from prompts or trending ideas, ready to upload in one workflow.


🚀Quick Shifts

📱 Apple is reportedly adding Grammarly-style AI writing assistance to iPhone and iPad, alongside AI-generated wallpapers, smarter Shortcuts automation, and a rebuilt Siri experience in upcoming updates.

⚡ A new U.S. bill from Sen. Adam Schiff aims to stop households from absorbing energy costs tied to AI data centers by placing stricter requirements on high-power AI infrastructure projects.

⚖️ Elon Musk lost his case against Sam Altman and OpenAI after a jury found key claims were blocked by statute-of-limitations rules, though Musk says he plans to appeal.

🎙️ Amazon’s Alexa Plus can now generate AI podcasts on nearly any topic, using synthetic hosts and news sources to create customizable audio explainers directly inside Alexa.

🏢 OpenAI partnered with Dell to bring Codex into corporate data centers, letting enterprises run AI coding agents closer to sensitive internal systems, repositories, and workflows in hybrid or on-premise environments. 

🧩 Prompt of the Day

Lookalike Audience Strategy That Finds Better Customers

Most lookalike audiences fail because the source audience is weak. High-performing campaigns start with the right customer signals and scale from proven buyers.

Turn your best customers into a scalable acquisition engine.

Paste the prompt: Drop this into ChatGPT, then fill in your audience and platform details.

Prompt to paste

Create a lookalike audience strategy for [Insert brand or product]. Identify the best source audience using data such as high-LTV customers, repeat buyers, or top converters. Recommend ideal audience sizes and expansion percentages for testing and scaling across platforms like Meta, Google, or TikTok. Analyze the key traits, behaviors, and interests shared by top customers, and explain how to structure campaigns around these segments to improve conversion rates and customer quality.

🤳AI Shift of the Day


That’s all for today’s edition. See you tomorrow as we track down and get you all that matters in the daily AI Shift!

If you loved this edition, let us know how much:

How useful was today's edition?

Login or Subscribe to participate in polls.

Forward it to your pal to give them a daily dose of the shift so they can 👇

Reply

or to participate.