{{first_name | Leader}}, welcome back. Today’s updates:
Learn what actually makes RAG work.*
Replit hits $3B as AI pair-programming drives growth
RSS co-creator launches new AI data licensing protocol
“Second-Brain” AI cuts search time 30%
Practical tools + must-read resources. Dive in ⬇️
Most RAG prototypes work in demos but fail with real users. This 6-week program shows you how to move past trial-and-error and build systems that deliver results in production.
What you'll learn:
✅ Embeddings & rerankers - When to fine-tune vs. when to use off-the-shelf
✅ Systematic evals - Build synthetic tests that predict production failures before deployment
✅ Error analysis - Identify exactly why retrievals fail and fix the root cause
✅ Online monitoring - Create feedback loops that automatically improve your system
✅ Agentic systems - Build RAG that routes queries intelligently and self-improves
✅ Production patterns - Transform simple Q&A into high-value report generation
What you’ll get:
6 hours of live instruction + 10 office hours with Jason Liu as your tech lead
12 Python notebooks with production-ready implementations
$2k+ in credits for Lance, Braintrust, ChromaDB, Cohere, Gemini, and more
Join alumni from Anthropic, OpenAI, Microsoft, Amazon, and top consulting firms
Special offer: Use code NEAT for 30% off - the largest discount offered so far.
It starts next week!
Replit Hits $3B as AI Pair-Programming Drives Growth
Replit now makes about $150M a year, and 40% of all coding on the platform runs through its AI helper, Ghostwriter. Enterprise customers have tripled in six months. The team also teased an “instant deploy” feature (think Copilot-style one-click release) coming in Q4 2025.
Low-code AI is taking over routine dev work and moving costs from salaries to pay-as-you-go usage. Set rules on which prototypes can launch straight from Replit. Compare Replit’s full build cost with traditional vendor statements of work before next year’s budget is locked.
RSS Co-Creator Launches New Protocol for AI Data Licensing
RSS co-creator Dave Winer has launched DSNP-AI, a blockchain tag that lets publishers mark, price, and track how AI models use their content. Five news sites and three data brokers start a public test next quarter. The system supports per-token fees and can auto-remove unlicensed copies.
If data turns into “streamed” media, expect new budget lines for content rights. List the outside data sources your models rely on now, and have procurement line up bulk DSNP deals before pay-per-token pricing begins
“Second-Brain” AI Cuts Search Time 30%
Three ex-Google X founders are building an assistant that pulls together calendars, email, and docs, then sends smart, just-in-time reminders. The invite-only beta already connects to Gmail, Slack, and Salesforce, with a SOC 2 audit due by year-end. Early users spend 30% less time hunting for information.
Workflows are shifting from “search and scroll” to AI that nudges you before you need it. Test the tool with one leadership team, track meeting-prep minutes saved, and roll out wider only if the savings beat your current knowledge-management spend.
Productivity Tools
HeyMilo – Phone-screen applicants with an AI recruiter that ranks fit and writes summaries.
TinyEinstein – Auto-build Shopify email flows and product promos that boost repeat purchases.
Twill – Offer employees digital health journeys that personalize care plans and track outcomes.
Helix – Chat-based IDE assistant that edits, debugs, and runs commands directly in your repo.
Get featured tomorrow: How do you use AI for business/personally? Interesting stories will be shared with 100K curious readers.
Useful Resources
30-day social calendar generator: reuse high-performing posts via plug-and-play ChatGPT prompts (link)
Satya Nadella’s 6 go-to prompts for lightning-fast executive decision memos (link)
Zapier audit flags rogue “shadow-AI” tools before they leak corporate data (link)
HubSpot’s 5-step checklist to lock AI drafts into your exact brand voice (link)
Slack backlog checklist that trims sprint overruns by 40% (link)
What’s the biggest benefit of using AI for predictions in business?
💰 Funding
Higgsfield raised $50M to generate Hollywood-grade AI video from text prompts.
Astrus secured $8M to speed analog-chip design with physics-aware AI.
💼 Who’s Hiring in AI
🐦 Learn Feel Lazy Again!
This guy literally explained why some people never feel lazy and how to copy them (in 2 mins)
— #Aadit Sheth (#@aaditsh)
12:56 AM • Sep 10, 2025
Creating Lineart in Real Photos
Add clean, minimal white line-drawing illustrations of people into this photo. Match the perspective, lighting, and scale of the scene. The illustrated figures should interact naturally and meaningfully with the environment, reflecting the mood, purpose, and activity of the space. Keep the drawings simple, fluid, and expressive, with no facial details. Maintain a modern, warm, and slightly whimsical tone that complements the overall aesthetic. Do not obscure any original elements. The illustrated figures should feel like friendly, imaginative additions that blend seamlessly with the context of the scene.
Attach ImageSample Output
P.S. Get more such prompts in the Prompting Playbook (free for you)
Q: When you hear AI, what’s the first company that comes to mind?
Poll Results: Over 60% people think of OpenAI when they hear AI.
Stay curious, {{first_name | leaders}}
PS. If you missed yesterday’s issue, you can find it here.