Good morning, here’s what you’ll learn today. We’re spotlighting TextJam, a writing tool that lets you sculpt text like clay instead of fighting with AI drafts. Plus, OpenAI’s deepfake crackdown after hollywood pushed back, Anthropic’s leap into drug discovery with Claude for Life Sciences, DeepSeek’s new OCR model that crushes 200,000 pages per day, and a $125 million unicorn bet on LangChain building the future of AI agents.
In Today's Edition
Today’s AI Tool Breakdown: TextJam
Quick overview
TextJam is a multiplayer writing app that blends human writing with AI help using a pen and pencil metaphor. Your pen marks lock in your voice, your pencil marks signal drafts the AI can refine. Think of it like Google Docs had a baby with ChatGPT that actually respects your style.
How to use it
Sign up at textjam.com and open a new document.
Write your solid ideas with the pen tool—anything in pen stays untouched.
Draft rough sections with the pencil tool—the AI reads these as fair game for edits.
Click the magic wand to let the AI refine pencil text while matching your pen style.
Drop in-place prompts (like sticky notes) next to text if you want the AI to make it shorter or add a fact.
Use pinch-to-resize gestures on your trackpad to shrink or expand sections—the AI rewrites to fit the new length instantly.
Copy/paste starter script
Open a blank doc, write your main point in pen, switch to pencil for rough supporting sentences, then hit the wand. The AI will fill gaps and match your tone without hijacking your voice.
Real-world use cases
Draft blog posts where you nail the intro in pen, rough out the body in pencil, and let AI polish the flow.
Co-write reports with teammates—each person picks their own AI model (Claude, GPT, Llama) on the same doc.
Write research papers by locking core findings in pen and prompting AI to expand methods in pencil.
Brainstorm emails in pencil, refine tone with prompts, then lock final version in pen.
Pro tips
Use the time traveler feature to rewind to any version or watch a sped-up video of your edits—great for seeing how AI changed your draft.
Switch AI models mid-document to compare outputs or get a second opinion from a different engine.
Dictate thoughts directly into the doc using the voice feature—it transcribes in real time.
Free vs paid
Free: Access to core pen/pencil tools, basic AI generations, and solo editing. Solid for testing the workflow.
Paid: Starter plan at $10/month gives 50,000 AI-generated words and audiobook-quality narration. Pro plan at $20/month bumps you to 200,000 words, full multiplayer editing, and dictation.
Alternatives
Notion AI — built into Notion, great for notes and wikis, but less precise control over what AI can touch.
Jasper — fast for marketing copy and ads, stronger templates, but less flexible for long-form collaboration.
Google Docs with Gemini — familiar interface, simpler AI sidebars, but no pen/pencil distinction or multi-model switching.
Quick demo video
https://www.youtube.com/watch?v=Ebea-a3qK8M
Today In AI News, The Top 4 Stories (And Why They Matter)
This matters because: Deepfake abuse threatens talent rights and trust in online video, forcing AI companies to balance innovation with consent and legal risk.
Quick summary: OpenAI announced new opt-in consent policies for Sora 2 after actor Bryan Cranston, SAG-AFTRA, and talent agencies complained about unauthorized AI-generated videos of public figures like Martin Luther King Jr. and Robin Williams. The company will block well-known characters and take down non-compliant content. Think of it like putting a permission gate on a deepfake factory.
This matters because: Drug research takes years and millions of dollars—AI that cuts data analysis from days to minutes could radically speed up new treatments and clinical trials.
Quick summary: Anthropic rolled out Claude for Life Sciences, connecting its Claude Sonnet 4.5 model directly to lab tools like Benchling and PubMed. Researchers can now pull data, analyze dosing plans, and auto-generate regulatory reports in minutes instead of days. Like hiring a research assistant who never sleeps and reads every paper instantly.
This matters because: Converting scanned docs, books, and handwritten notes into searchable text at scale unlocks huge archives for AI training and enterprise search.
Quick summary: DeepSeek-AI launched DeepSeek-OCR, a 3 billion parameter model that compresses text into 2D vision tokens and achieves 97% accuracy at 10× compression. It handles over 200,000 pages daily on a single A100 GPU and beats common baselines on OmniDocBench while using far fewer tokens. Imagine turning a library into a spreadsheet overnight.
This matters because: LangChain builds the plumbing that lets AI agents talk to apps and databases—if agents take off, this infrastructure becomes as critical as cloud storage.
Quick summary: LangChain raised $125 million in a round led by IVP, pushing its valuation past $1 billion. The platform powers AI agents for enterprise customers, helping them automate workflows that span multiple tools and data sources. It's like AWS for the agent economy—investors bet everyone will need it.
Thats All For Today!
For all questions, comments, concerns, or if you want us include anything specific - feel free to reply to this email! We will answer 😄