💻 Cursor built its own AI coding model

Hey folks,

Nvidia just became the first company to hit a $5 trillion market cap.. They crossed $4 trillion just three months ago. In May 2023, they were at $1 trillion. That's a 5x jump in 29 months, driven entirely by AI chip demand. The stock's up 1,500% over five years. No company has ever grown this fast to this scale. The AI infrastructure buildout is accelerating fast!

Let's dive in..

💻 Cursor 2.0 Ships Its Own AI Coding Model

🐝 The Buzz: Cursor launched Composer, its first in-house AI model built specifically for software engineering, alongside Cursor 2.0's multi-agent architecture. Instead of relying on GPT or Claude, Cursor trained a mixture-of-experts model using reinforcement learning across hundreds of thousands of sandboxed coding environments. The model generates code at 250 tokens per second, 4x faster than competing tools.

• Reinforcement learning training: Composer learned from real-world software engineering challenges, optimized for tasks like editing code, fixing linter errors, running terminal commands, and writing unit tests.

• 250 tokens/second generation: Uses MXFP8 low-precision training for fast inference while maintaining frontier-level coding intelligence.

• Multi-agent architecture: Developers can launch eight parallel agents, each operating independently with semantic search, terminal access, and plan generation capabilities.

💡 Takeaway: Cursor's vertical integration into model training signals that AI coding tools need purpose-built models, not general-purpose LLMs. For engineering teams, this means faster iteration cycles and better tooling designed specifically for code not adapted from chatbots.

Together with The Deep View

Stop Drowning In AI Information Overload

Your inbox is flooded with newsletters. Your feed is chaos. Somewhere in that noise are the insights that could transform your work—but who has time to find them?

The Deep View solves this. We read everything, analyze what matters, and deliver only the intelligence you need. No duplicate stories, no filler content, no wasted time. Just the essential AI developments that impact your industry, explained clearly and concisely.

Replace hours of scattered reading with five focused minutes. While others scramble to keep up, you'll stay ahead of developments that matter. 600,000+ professionals at top companies have already made this switch.

Join them today, for free.

🎨 Adobe Firefly Goes All-In on Multi-Model AI

🐝 The Buzz: Adobe announced Firefly Image Model 5 generating native 4MP photorealistic images with prompt-based editing in everyday language, plus timeline-based AI video editor with Generate Soundtrack and Generate Speech. The strategic shift: Photoshop's Generative Fill now supports third-party models like Google Gemini 2.5 Flash, Black Forest Labs FLUX.1 Kontext, and Firefly Model 5, giving creators multiple variations without leaving the app.

• Third-party model integration in Photoshop: Generative Fill users can now choose between Firefly, Gemini 2.5 Flash, FLUX.1 Kontext, or custom models, all within the same workflow for creative flexibility.

• Firefly Image Model 5: Generates studio-quality 4MP resolution images and handles prompt-based editing in conversational language, while Layered Image Editing ensures realistic shadow and light handling.

• AI-driven audio/video tools: Generate Soundtrack creates fully licensed instrumental tracks, Generate Speech produces crystal-clear voiceovers with ElevenLabs integration, and new timeline-based video editor handles multi-track editing.

💡 Takeaway: Adobe's embracing multi-model orchestration over proprietary lock-in. For engineering teams building creative pipelines, this validates the best-of-breed approach, let users choose models per task instead of forcing single-vendor toolchains. Premiere Pro's AI Object Mask shows how automation moves beyond image generation into video production workflows.

📰 Like newsletters? Here are more our readers enjoy.

💻️ How to Permanently Remove — Em Dashes

You’ve probably noticed that when people talk about AI writing, the dreaded em dash (“—”) keeps getting mentioned. What used to be the stylish pause — is now seen by many as the tell-tale sign of “AI wrote this.” But instead of just asking ChatGPT to “stop using em-dashes”, there’s a smarter fix: teach it what to use instead. Thanks to Ruben Hassid’s guide on style-training AI, we can see how to alter instructions, and fix the —.

• Go into ChatGPT’s Custom Instructions 
  (bottom-left menu → Personalization → Custom Instructions).

• Under Custom Instructions, add the following

“Systematically replace em-dashes (“—”) with a period (“.”) to start a new sentence, or a comma (“,”) to continue the sentence.”

• Save the instructions. That way you’re giving ChatGPT what to do, not just what not to do.

🐝 AI buzz bits

✍️ Grammarly rebranded as Superhuman and launched Superhuman Go, an AI assistant built into the Grammarly browser extension. It connects to Jira, Gmail, Google Drive, and Calendar to log tickets, fetch meeting availability, and retrieve data from CRMs and internal systems.

🎬 Synthesia raised $200M at a $4B valuation just a week after rejecting Adobe's $3B takeover offer. GV (Google Ventures) led the round for the AI video platform that turns text into videos with lifelike avatars speaking in multiple languages. This doubles their January 2025 valuation of $2.1B and brings total funding to $536M.

🛠️ Microsoft responded to OpenAI's Atlas browser within 48 hours. The update added Actions (automated form filling and hotel booking) and Journeys (tab connection tracking) to Copilot Mode. It's the clearest signal yet that Microsoft sees AI browsing as strategic territory worth defending fast.

🔥 AI productivity tools

For a full list of 1000+ AI tools, visit our Supertool Directory

🤘 TGIF