- AI Valley
- Posts
- OpenAI plans fully automated AI researchers by 2028
OpenAI plans fully automated AI researchers by 2028
PLUS: OpenAI launches Aardvark, an AI security researcher
Together with
Howdy, it’s Barsee.
Happy Friday, AI family, and welcome to another AI Valley edition. Not much happened today, so this issue takes just 3 minutes to read.
Today’s climb through the Valley reveals:
OpenAI plans fully automated AI researchers by 2028
OpenAI launches Aardvark, an AI security researcher
Big tech is spending more than ever on AI and it’s still not enough
Plus trending AI tools, posts, and resources
Let’s dive into the Valley of AI…
MINIMAX
8% Claude Sonnet price, ~2x faster
Advanced Coding Capability: Engineered for end-to-end developer workflows. Strong capability on a wide-range of applications (Claude Code, Cursor, Cline, Kilo Code, Droid, etc)
High Agentic Performance: Robust handling of long-horizon toolchains (mcp, shell, browser, retrieval, code)
#1 Open-source Model & #4 Intelligence & #1 Token Usage among open-source models on OpenRouter
Use it now:
2️⃣ Access M2 API
3️⃣ Open-sourced for local deployment.
*This is sponsored
THROUGH THE VALLEY
OpenAI CEO Sam Altman said the company aims to build an AI “research intern” by 2026 and a fully autonomous “AI researcher” by 2028. Joined by Chief Scientist Jakub Pachocki, Altman described a system that can plan, run experiments, and make discoveries on its own. To get there, OpenAI is improving algorithms, scaling compute, and extending model reasoning time. The announcement came as OpenAI completed its shift to a public benefit corporation, with a $25 billion pledge toward disease research and a $1.4 trillion infrastructure plan.
OpenAI has launched Aardvark, an autonomous AI security agent built on GPT-5 that scans codebases to find, confirm, and fix vulnerabilities. Now in private beta, Aardvark reviews repositories, tracks commits, and suggests patches using reasoning instead of traditional tools like fuzzing. In tests, it identified about 92% of known and synthetic bugs across enterprise and open-source projects. The tool integrates with GitHub and Codex for one-click patching and annotated fixes, signaling a move toward AI systems that actively protect software as it evolves.
Big Tech plans to spend over $400 billion on AI this year and says it’s still not enough. Microsoft will double its data center capacity within two years as demand surges, while Amazon races to add cloud infrastructure. Meta says it’s hitting limits training new models and will sharply increase spending again in 2026. Google is raising annual capital spending to around $93 billion, citing strong AI returns. The companies see massive opportunity in AI but face growing investor concern over whether all this spending will pay off.
TRENDING TOOLS
Superpower > Detect early signs of 1,000+ health conditions
Sora App Cameos > OpenAI’s Sora now lets you bring back your favorite characters: your dog, toys, or wildest creations for quick cameos
Skyvern > Automate anything on your browser
Canva Design Model > A new AI model trained to grasp the full complexity of design, from color harmony to layout balance and creative intent
Devin > It now has full computer use capabilities and can share screen recordings
Kimi K2 > The open-source trillion-parameter MoE AI model for advanced coding assistance, intelligent agents, and automated workflows
OpenAI Aardvark > A GPT-5–powered security researcher that autonomously finds, analyzes, and reports vulnerabilities. Currently in private beta
Microsoft Researcher > With Computer Use, it can now explore both the open and gated web, connect insights across hundreds of sites, and generate rich, multi-step reports
Kaizen > Build browser automations that handle repetitive data entry once and for all, so you never do the same task twice
THINK PIECES / BRAIN BOOST
THE VALLEY GEMS
What’s trending on social today:
New Anthropic research: Signs of introspection in LLMs.
Can language models recognize their own internal thoughts? Or do they just make up plausible answers when asked about them? We found evidence for genuine—though limited—introspective capabilities in Claude.
— Anthropic (@AnthropicAI)
5:18 PM • Oct 29, 2025
THAT’S ALL FOR TODAY
Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee
👍️ New reader? Subscribe here
Thanks for being here.
HOW WAS TODAY'S NEWSLETTER |
REACH 100K+ READERS
Acquire new customers and drive revenue by partnering with us
Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.
If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.



