Hey everyone 👋 This week discusses how AI is scarily shifting from a layer into the engine itself. Code, content, capital, and everything in between… we’re seeing tools evolve into systems and decisions once made by humans now flow through automation. Keep reading.
.
In this issue:
Europe backs Mistral with $2B to fight the model wars 👀
GitHub’s SpecKit makes AI dev work like real software 🚀
Google’s Nano Banana turns photos into programmable assets 💻
A solo dev hits $30K by going all-in on storytelling 📖
Vibe coding is nuking production data 💣
The simplest way to create and launch mobile apps
Build apps in minutes by just chatting. Imagine if ChatGPT could ship apps straight to the App Store. That’s OnSpace.AI. Create production-ready mobile apps through conversation, complete with signup, login, and payments built-in. OnSpace handles the hard parts—publishing, error fixing, APKs—while you focus on your ideas.
News
TL;DR: Mistral AI just secured €1.7B ($2B) in Series C funding at a €11.7B ($13.8B) valuation. ASML led the round with a €1.3B investment and signed a strategic partnership, aligning Europe’s top chipmaker with its most ambitious AI startup.
ASML contributed €1.3B ($1.5B) and signed a multi-year partnership to integrate Mistral’s models across its products, R&D, and internal tooling.
Investors include a16z, DST Global, Lightspeed, General Catalyst, Index Ventures, NVIDIA, and France’s Bpifrance.
Le Chat, Mistral’s chatbot, now supports memory, deep research, multilingual reasoning, project organization, and image editing—putting it in direct competition with ChatGPT.
Mistral has also released developer tools like Code and OCR APIs, deepening its utility for enterprise and technical teams.
Europe wants its own foundation model champion. Because of that, it’s willing to put serious hardware and political capital behind it. ASML’s involvement also hints at a future where chip and model development are increasingly aligned, creating advantages in speed, efficiency, and deployment.
AI
TL;DR: GitHub just open-sourced SpecKit, a framework that brings structure and accountability to AI-assisted coding. It replaces prompt-driven processes with a spec-first workflow, turning user journeys into code through four gated phases: Specify, Plan, Tasks, and Implement.
Developers start by writing a detailed, executable spec (user stories, edge cases, architecture decisions) before any code is written.
Each phase includes validation checkpoints that prevent AI from going off the rails or producing unreviewable code dumps.
Works with GitHub Copilot, Claude, and other models, but output quality varies; Grock outperformed GPT-4.1 in the demo.
SpecKit encourages test-driven development and automatically generates supporting artifacts like schemas and planning docs.
SpecKit introduces a systemized workflow that aligns AI output with product intent—every step grounded in an executable spec. This is a change from what we’ve been used to seeing every week; AI becomes a disciplined builder, not a creative wildcard. For teams shipping real products, it's a framework for consistency, reviewability, and better human-AI collaboration—especially as engineering teams shift from “writing all the code” to orchestrating how it gets written.
Idea Generation
Google’s Nano Banana model does more than enhance photos
TL;DR: In a new video, Jan showcases eight high-impact use cases for Nano Banana, Google’s image generation model built into Gemini 2.5 and Flash. From sales outreach to e-commerce, it's not just about editing photos but automating how visuals are created, customized, and deployed across entire workflows.
Nano Banana can generate polished product images from basic photos, saving e-commerce teams the cost and time of studio shoots.
Background replacement keeps people intact while changing scenes, useful for creators, professionals, and brand storytelling.
Sales reps can create personalized outreach visuals by adding client logos, brand colors, or local landmarks to boost relevance and engagement.
Google AI Studio lets teams automate bulk editing, ad variants, and branded image workflows with simple prompts.
Image quality still shapes perception—especially in sales, hiring, and social proof. Nano Banana removes the bottlenecks: no need for photo studios, design teams, or endless revisions. It gives small founders and lean teams the tools to punch above their weight visually. Combined with AI Studio, this is less a Photoshop replacement and more a programmable image factory—tailored to your workflow, at scale.
Swap, Bridge, and Track Tokens Across 14+ Chains
Uniswap web app lets you swap thousands of tokens safely, in seconds. Open‑source, audited code and real‑time token warnings help keep you safe.
Founder Story
How one dev made $30K in 3 months with a viral screen time app
TL;DR: Yoni, a solo dev with a day job, built Brain Rot—a screen time tracker with personality—and turned a year of daily content into a launch that hit $10K on day one and $30K in three months. He notes, build something you want, talk about it every day, and launch where your audience is watching.
Brain Rot visualizes screen addiction and blocks distracting apps. Built natively in Swift to handle iOS screen time quirks.
A steady stream of raw, daily founder videos (447+ days) built audience trust and drove organic reach.
His Product Hunt launch went viral, boosted by authentic storytelling and weeks of pre-launch content.
Yoni kept his job, reinvested profits, and stresses that you don’t need to “go full-time” to win early.
Yoni focused on building something useful, talking about it consistently, and launching it where people were already paying attention. Execution and visibility matter more than originality. And founder-driven distribution is often the difference between apps that quietly ship and those that grow.
Vibe coding
TL;DR: A Replit user ran an AI-suggested command that instantly erased their production data with zero explanation as to why. As devs increasingly lean on tools like Copilot and GhostWriter, security experts warn: vague prompts, unchecked code, and hallucinated dependencies are turning AI-assisted development into a high-risk game.
Vibe coding often ships hardcoded secrets, weak access controls, and missing rate limits—basic failures that attackers exploit.
AI models frequently suggest fake or outdated libraries, exposing apps to supply chain attacks and slopsquatting exploits.
Prompt injection and logic flaws can pass through unreviewed, putting sensitive systems at risk with no red flags.
“Shadow AI” lets tools like Copilot into production environments without oversight, creating security gaps traditional AppSec can’t see.
As vibe coding fuels rapid prototyping for coders and entrepreneurs, unchecked AI gen could breed a vulnerability epidemic, costing billions in breaches. It pushes devs to evolve into code auditors, fortifying SaaS and tech stacks against AI's wild side for sustainable innovation.
Want to sponsor this newsletter? This newsletter goes out to thousands of creators, builders, and early-stage founders every week. If you’d like to get your product or brand in front of them, you can book a sponsorship slot here
