- AI Shots
- Posts
- 💻 AI Coding? Not So Smart
💻 AI Coding? Not So Smart
PLUS: 🎨 YouTube Turns Photos Alive!

Hey AI Explorers,
Here’s what’s in store for you today:
💻 AI Coding Gets a Reality Check — With Just 7.5% Accuracy
🎨 YouTube Shorts Adds AI Video Tools — Turn Your Photos into Moving Stories
🇺🇸 Trump’s AI Action Plan: Build Fast, Regulate Less, Compete Hard
LATEST DEVELOPMENT
💻 AI Coding Gets a Reality Check — With Just 7.5% Accuracy

💻 AI Coding Gets a Reality Check — With Just 7.5% Accuracy
The first round of the K Prize, a new AI coding challenge launched by Databricks and Perplexity co-founder Andy Konwinski, now has a winner — and the results are raising eyebrows across the AI world.
Eduardo Rocha de Andrade, a prompt engineer from Brazil, took home the $50,000 prize, but his winning score? Just 7.5% of the problems solved correctly
👉 That’s not a typo — and that’s the point.
While benchmarks like SWE-Bench have shown models hitting 75%+ on simpler tasks, the K Prize was built to be contamination-free, meaning no pretraining or exposure to the test set. It uses fresh GitHub issues that were flagged after model submissions closed, forcing models to truly generalize.
“Benchmarks should be hard if they’re going to matter,” Konwinski said.
He’s now offering $1M to the first open-source model that scores above 90% on this challenge — a major incentive to push real progress, not leaderboard fluff.
AI coding agents are often hyped as near-ready replacements for junior developers. But if top models can’t even clear 10% on a fresh benchmark, we may be further from AI software engineers than the hype suggests.
Expect the next round of the K Prize to drop in the coming months. For now, it’s a reminder that real-world AI benchmarks need to evolve — because solving GitHub tickets isn’t as easy as it looks in a demo.
🎨 YouTube Shorts Adds AI Video Tools — Turn Your Photos into Moving Stories

YouTube is supercharging Shorts with new generative AI features, giving creators more ways to bring static content to life.
Rolling out this week in the U.S., Canada, Australia, and New Zealand, the new tools include:
➡️ Image-to-Video AI: Upload a photo, and YouTube turns it into a 6-second animated video — zooming, moving, or even morphing elements into motion.
➡️ AI Effects: Turn doodles into digital art or transform your selfie into a scene — like swimming underwater or twinning with someone else.
These features are powered by Google’s Veo 2 model and come with clear AI labels and SynthID watermarks for transparency.
You can try them by heading to the Shorts camera, tapping the “Effects” icon, then selecting “AI.” YouTube has also introduced AI Playground, a new hub for experimenting with AI tools and prompts — now live in supported regions.
📈 Why it matters:
With 200B+ Shorts views daily, YouTube is pushing hard to keep creators ahead of the curve as AI-driven creativity becomes the norm. This rollout is a warm-up — the more powerful Veo 3, which adds sound and longer video generation, is coming to Shorts later this summer.
🇺🇸 Trump’s AI Action Plan: Build Fast, Regulate Less, Compete Hard

The Trump administration has released its AI Action Plan, taking a sharp turn from Biden’s cautious approach. The focus? Infrastructure, deregulation, national security, and open-source AI.
Key highlights:
🏗️ Data center boom: Plan prioritizes building AI infrastructure — even on federal lands and during peak energy demands. Environmental rules may be bypassed to “Build, Baby, Build.”
🧯 Less regulation: Push to weaken state-level AI laws by tying compliance to federal funding. FCC may intervene where state rules conflict with federal AI goals.
🧠 National security first: AI is now core to defense strategy. DoD will get dedicated compute, military AI training, and intelligence on adversary AI systems.
🗣️ Speech over safety: DEI, misinformation, and climate language will be stripped from federal AI policy. Government contracts will favor “neutral” models — though “neutrality” remains undefined.
🔓 Open-source boost: Plan backs open AI labs like Meta, Hugging Face, and AI2, with funding and access to compute to counter China’s rise in open AI.
🔐 Safety lite: Includes hackathons and research on robustness, but skips mandatory safety reporting seen in Biden’s order.
📌 Bottom line: Trump’s AI vision is clear — go big, go fast, and treat AI like a race America must win. Ethics, regulation, and environmental safeguards take a backseat to speed and dominance.
QUICK HITS
📰 Everything else in AI today
🗣️ ElevenLabs drops Conversational AI 2.0
🧠 OpenAI teases “ambient” hardware devices
💰 Anthropic hits $3B annualized revenue
🔐 Meta automating 90% safety reviews
🎥 Veo 3 used in millions of videos
Whenever you're ready, here are ways we can support each other:
Promote your product or service to 100K+ global professionals, AI enthusiasts, entrepreneurs, creators, and founders. [Contact us at [email protected]]
Refer us to your friends and colleagues to help them stay ahead in the latest AI developments. We've helped 30K+ creators, entrepreneurs, founders, executives, and others like you.
That's it for today!Before you go we’d love to know what you thought of today's newsletter to help us improve The AI Shots experience for you. |