Hey! Glad you’re here. I started this newsletter because I found myself spending hours every week digging through product launches, research threads, and tool directories to stay on top of what was actually worth paying attention to. There’s so much noise in AI right now that the signal gets buried fast. So I built a system to surface it and decided to share it. Every issue, I personally review what shipped, filter out the hype, and drop only what I’d actually use or recommend. No sponsored rankings, no filler. Just the tools that made me stop scrolling. Thanks for reading.
This week’s batch is all about tools that do the work while you sleep. Digital twins, agentic accountants, and open-source video studios running on your own GPU.
✍️ Writing & Content
Friendware — System-wide tab-to-complete for macOS that reads your screen context and writes inline suggestions in any app.Try this: CC a coworker on Slack, hit Tab, and let it draft your reply using the thread as context.
Paid → Try it
🎨 Image & Design
Nano Banana 2 — Google’s upgraded image generation engine now baked into Gemini and Lens, delivering faster output and significantly improved photorealism.Try this: Open the Gemini app and ask it to generate a product mockup the jump in realism from v1 is immediately noticeable.
Free → Try it
🎬 Video & Audio
Seedance 2.0 — ByteDance’s multimodal video generator that accepts images, video clips, audio, and text as input and generates synchronized dialogue and sound effects alongside the visuals.Try this: Upload a product photo + a 10-second voiceover, and get a polished promo clip with matching ambient audio.
Freemium → Try it
LTX Desktop — A fully open-source, local-first video editor with built-in AI generation (text-to-video, image-to-video, retakes) powered by the LTX-2.3 engine. No cloud, no per-generation fees.Try this: Import a rough timeline from Premiere, then use AI retakes to regenerate weak shots entirely on your own GPU.
Free / Open Source → Try it
💻 Coding & Dev
Codex Security — OpenAI’s new security auditor that builds deep context about your codebase to surface complex vulnerabilities other scanners miss, with 90% fewer false positives.Try this: Point it at a production repo and compare its prioritized report to your existing SAST output the signal-to-noise difference is stark.
Paid → Try it
Codex Desktop App — OpenAI’s new standalone app for managing AI coding agents and automating multi-step developer workflows beyond the browser.Try this: Spin up an agent to handle a full PR cycle from branch creation to test runs while you focus on architecture.
Paid → Try it
📊 Productivity & Ops
Read AI — Ada — A digital twin you activate by CC’ing ada@read.ai on any email thread. It schedules meetings, answers questions from your knowledge base, and handles follow-ups autonomously.Try this: CC Ada on a scheduling thread it reads everyone’s calendars and proposes times without you lifting a finger.
Free → Try it
Simplora 2.0 — An agentic meeting stack that preps you before calls, surfaces real-time answers during them, and generates role-adapted action items afterward now with a desktop app that skips the meeting bot entirely.Try this: Let it join your next Zoom after the call, check the personalized notes adapted to your specific role on the team.
Freemium → Try it
Basis — An agentic accounting platform that automates complex bookkeeping workflows end-to-end, designed for startups drowning in spreadsheet-based finance ops.Try this: Connect your bank feeds and let it categorize, reconcile, and flag anomalies then review instead of doing.
Paid → Try it
🤝 Sales & Marketing
Monaco — An AI-native CRM from ex-Founders Fund VC Sam Blond, built from scratch for seed/Series A startups with a built-in ZoomInfo-style prospecting database. $35M raised.Try this: Import your ICP criteria and let Monaco auto-build a prospecting list with verified contacts no separate data vendor needed.
Paid → Try it
🔍 Research & Knowledge
DeepSeek V4 — A 1-trillion-parameter open model that activates only 32B parameters per token, delivering frontier-class reasoning at a fraction of the compute cost.Try this: Run it locally via Ollama for deep research tasks where you need powerful reasoning without sending data to the cloud.
Free / Open Source → Try it
🤖 Agents & Automation
Gemini 3.1 Flash-Lite — Google’s ultra-light, ultra-fast variant of Gemini 3, purpose-built for high-volume agentic workloads at rock-bottom cost.Try this: Use it as the backbone for a customer-support agent pipeline where you need sub-second latency on thousands of concurrent queries.
Freemium → Try it
AI Headlines
OpenAI ships GPT-5.4 Thinking with 1M-token context and 33% fewer factual errors → link
95% of developers now use AI tools weekly; Claude Code is the most-used coding agent → link
Apple’s reimagined AI Siri with on-screen awareness targets March iOS 26.4 release → link
Tencent open-sources HunyuanVideo RL post-training code for real-time interactive world models → link
Gartner projects worldwide AI spending to hit $2.52 trillion in 2026 → link
Tool of the Drop
Read AI — Ada stands out this issue because it solves one of the most universal productivity drains scheduling and follow-up without requiring you to learn a new app or change any workflow. You just CC an email address. The fact that it draws on your meetings, files, and connected systems to answer real questions (not just find calendar slots) makes it feel less like a bot and more like an actual assistant. And it launched free to all 5 million Read AI users on day one.