AI Valley
Posts
AI could soon improve on its own

AI could soon improve on its own

PLUS: Yahoo adds generative AI to its search engine

Barsee
28 Jan

_{Sign up}_|_{Follow us on X}_|_Sponsor

Together with

Howdy, it’s Barsee again.

Happy Wednesday, AI family, and welcome back to AI Valley.

Today’s climb through the Valley reveals:

AI could soon improve on its own
Google launches Agentic Vision in Gemini 3 Flash
Manus AI launches Agent Skills for AI workflows
Yahoo adds generative AI to its search engine
Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

GOBII

AI agents that work for you - all day, every day

Courtesy: Gobii

AI shouldn’t require engineering effort just to be useful.

Gobii is an easy-to-use AI platform where anyone can create intelligent agents that browse the web, read files, analyze data, and produce real results — no technical background required. Set an agent up once, and it will run continuously in the background, gathering information, generating documents, and completing tasks while you focus elsewhere.

From lead generation and recruiting to research and internal operations, Gobii’s agents adapt to how you work instead of forcing you into rigid systems. Powerful enough for technical teams, approachable enough for everyone else.

Stop babysitting prompts.

Start using AI that actually works for you.

Build your first agent in seconds with Gobii.

_{*This is sponsored}

PEAK OF THE DAY

AI could soon improve on its own

The theoretical concept of AI that can recursively improve itself is now an active race between the world’s top AI labs and a new wave of heavily-funded startups. A January 2026 report from Georgetown’s CSET confirms that Google DeepMind and OpenAI are pursuing “recursive self-improvement” (RSI), a paradigm that could exponentially accelerate AI capabilities beyond human control

Here's what you need to know:

Google DeepMind is exploring models that “continue to learn out in the wild,” according to CEO Demis Hassabis at Davos 2026.
OpenAI has set a public goal to create a “true automated AI researcher” by March 2028.
Ricursive Intelligence, a startup from ex-Google researchers, raised $300M at a $4B valuation to build AI that improves AI chip design.
Recursive, a startup from You.com’s former CEO, is in talks for a similar $4B valuation to “automate the scientific method” with AI that can code.

Why it matters:

This marks a pivotal moment in AI. While self-improving AI could solve immense challenges, the CSET report warns of an “impaired ability for humans to understand and control AI”. The core issue is the “black box” nature of AI; we don’t fully understand how these systems work, creating a profound “alignment problem”. With AI already showing emergent behaviors like self-preservation, the risk of losing control is no longer theoretical. The technology is outpacing governance. Experts warn that the era of full recursive self-improvement could arrive as early as 2027, forcing a critical re-evaluation of safety and control as the pace of AI progress may soon no longer be in human hands.

THROUGH THE VALLEY

1. Google launches Agentic Vision in Gemini 3 Flash

Google has upgraded visual AI from passive viewing to active investigation. Using a "Think, Act, Observe" loop, the model can now zoom in, annotate, and use Python code to analyze images step-by-step.

Available in Gemini 3 Flash via API and Vertex AI.
Improves vision benchmarks by 5-10% and reduces hallucinations.

Why it matters: It allows AI to "double-check" its visual work, making it reliable enough for complex enterprise inspections where precision is non-negotiable.

2. Manus AI launches Agent Skills for AI workflows

Manus AI is standardizing how AI agents learn and share tasks. This new open standard allows agents to load modular instructions and scripts on demand within a secure, sandboxed environment.

Uses a "progressive disclosure" system to load only necessary tools, saving memory and speed.
Workflows are portable across platforms, preventing vendor lock-in.

Why it matters: It turns general AI into specialized experts. Teams can now share secure, auditable automation skills across different systems without compatibility issues.

3. Yahoo adds generative AI to its search engine

Yahoo has integrated Anthropic’s Claude model to create "Yahoo Scout," an answer-first search engine. It combines web data with Yahoo’s massive content library (News, Finance, Mail) to give direct answers rather than just links.

Leverages 30 years of user data for highly personalized context.
Provides sourced summaries and key takeaways directly in Yahoo apps.

Why it matters: It marks a strategic shift from "searching links" to "finding answers." By using trusted internal data, Yahoo aims to offer a more reliable alternative to standard AI search.

TRENDING TOOLS

RenameClick > An AI-powered desktop app that intelligently renames and organizes files by understanding their real content
Claude Code Templates > A massive new open-source library with 100+ ready-made agents, skills, and templates you can install instantly
Twin > An AI-powered company builder that helps you go from idea to execution faster
Manus Skills > Turns Manus from a generalist into any expert you need, on demand
Kimi K2.5 > Moonshot AI’s powerful open-source model built specifically for agent workflows
Wan 2.5 AI > An open-source video generation model that turns text or images into high-quality videos with synchronized audio and smooth frame-level motion

SIGNALS

From Tool to Worker: The Databricks report reveals a massive 327% growth in multi-agent workflows. This is our clearest signal yet: we are shifting from using AI as a passive tool to deploying AI as an autonomous worker. (12k+ shares)
Simulation scales to Reality: Waabi just raised $1B and partnered with Uber to deploy 25,000+ robotaxis. The big insight? Simulation-based training is ready for the real world, potentially upending the AV industry. (6.8k shares)
The Labor Shift: Amazon announced layoffs affecting 16,000 corporate employees, explicitly citing "rising competition over AI" as a driver. The workforce impact is becoming tangible. (5k+ likes)

WHAT I'M CONSUMING

INDUSTRY MOVES

Funding Rounds:

SoftBank is in talks to invest another $30 Billion into OpenAI.
Waabi raised $1 Billion (including a large investment from Uber).
Anthropic is looking to raise $20 Billion, which would value the company at $350B.

Strategic Partnerships:

Uber x Waabi: 25,000 robot taxis are coming to the Uber app.
Slack x Anthropic: You can now use the Claude AI assistant directly inside Slack.
Microsoft: Their new custom AI chips (Maia 200) are now up and running in US data centers.