• The AI Pulse
  • Posts
  • šŸ¤– Fake Software Company Run Entirely With AI Agents

šŸ¤– Fake Software Company Run Entirely With AI Agents

PLUS: Google’s Gemini 2.5 Pro Plays Pokémon Blue!

Welcome back AI enthusiasts!

In today’s Daily Report:

  • šŸ‘”Fake Software Company Run Entirely With AI Agents

  • šŸ•¹ļøGoogle’s Gemini 2.5 Pro Plays PokĆ©mon Blue!

  • šŸ› Trending Tools

  • 🄪Brief Bites

  • šŸ’°Funding Frontlines

  • šŸ’¼Who’s Hiring?

Read Time: 3 minutes

šŸ—žRECENT NEWS

AI RESEARCH

šŸ‘”Fake Software Company Run Entirely With AI Agents

Image Source: Canva’s AI Image Generators/Magic Media

Researchers at Carnegie Mellon University (CMO) developed a fake software company run entirely with AI Agents, and it was a total disaster.

Key Details:
  • We interact with computers on a daily basis, and many aspects of our everyday workflows depend on them. What if computers didn’t need us to operate them? That’s where AI Agents come in.

  • AI Agents are like virtual employees who can autonomously plan, execute, and refine their actions. So, what does that mean?

  • Unlike rule-based programs, like a Spam Filter that relies on a predefined set of rules to classify incoming emails as either ā€œSpamā€ or ā€œNot Spam,ā€ AI Agents analyze digital environments in real-time to plan, execute, and adapt.

  • Anthropic just started releasing AI Agents. For example, Computer Use (CU) uses a computer just like we do. But how effective is CU? Can it replace you?

  • TheAgentCompany was staffed with AI Agents from Meta, Google, OpenAI, and Anthropic. These AI Agents were plagued with a lack of common sense, weak social skills, and a poor understanding of how to navigate websites.

Why It’s Important:
  • The best-performing AI Agent was powered by Anthropic’s Claude 3.5 Sonnet, which still struggled to finish just 24% of the assigned workflows.

  • Even this poor performance was shockingly expensive, costing $6.10 per step with an average of 30 steps per workflow.

🩺 PULSE CHECK

What do you believe is the main bottleneck for adopting Agentic AI?

Vote Below to View Live Results

Login or Subscribe to participate in polls.

GOOGLE

šŸ•¹ļøGoogle’s Gemini 2.5 Pro Plays PokĆ©mon Blue!

Image Source: Twitch/{Gemini_Plays_PokĆ©mon}/ā€œTesting Gemini 2.5 Pro to see how well it plays PokĆ©mon Blue!ā€/ā€œImproved memory management?ā€/Screenshot

Google CEO Sundar Pichai recently shared that Gemini 2.5 Pro completed PokƩmon Blue.

Key Details:
  • Gemini 2.5 Pro is a multimodal AI model with reasoning capabilities across text, images, audio, video, and code. It’s designed to handle more complex problems that require multi-step reasoning.

  • It’s equipped with a hybrid reasoning engine that allows it to mimic a human’s decision-making capabilities and problem-solving capacity.

  • It leverages Chain-of-Though (CoT) techniques to break down complex problems into manageable sub-problems. Then, it solves each manageable sub-problem, combining them into a complete answer.

  • Gemini 2.5 Pro constantly captures the Game State: a current ā€œsnapshotā€ of the video game. Then, it generates a sequence of actions based on that current ā€œsnapshotā€ through commands like A, B, Up, Down, Left, Right, Start, and Select.

Why It’s Important:
  • Playing PokĆ©mon Blue requires not only understanding the video game’s mechanics but also applying decision-making to craft strategies that adapt to the ever-changing Game State.

  • The ability of Gemini 2.5 Pro to continuously and effectively interpret the ever-changing Game State offers a compelling parallel to tackling intricate, real-world tasks.

šŸ› TRENDING TOOLS

šŸ–Œļøremovebg removes image backgrounds for free.

šŸ“ŠFortune automates, analyzes, and forecasts your finances.

šŸ“ˆfreebeat. turns music and ideas into viral videos in seconds.

šŸ“¦Spring.new creates tailor-made business applications instantly.

āš™ļøomi connects 100+ apps to ChatGPT to make it 5x more personal.

🧰 Browse our Always Up-To-Date AI Toolkit.

🄪BRIEF BITES

Google Labs launched ā€œLittle Language Lessons,ā€ a collection of bite-sized learning experiments built with Gemini.

Apple and Anthropic are teaming up to build an AI-powered coding tool that writes, edits, and tests code for programmers.

Figure AI has sent legal notices to at least two brokers who run secondary marketplaces demanding they stop marketing the company’s shares.

Suno introduced ā€œv4.5,ā€ a music generator with expanded genres, smarter mashups, more textured sounds, extended song lengths, and richer vocals.

šŸ’°FUNDING FRONTLINES

  • Cluely closes a $5.3M Seed Round for invisible AI to cheat on meetings, conversations, or sales calls.

  • Ascertain raises a $10M Series A to reimagine operational workflows with Agentic AI.

  • Astronomer secures a $93M Series D to build, run, and observe data pipelines.

šŸ’¼WHO’S HIRING?

  • Conductor (New York City, NYC): AI Engineering Intern, Summer 2025

  • FanDuel (Atlanta, GA): Data Analyst, Entry-Level

  • Glassdoor (Austin, TX): Social Media Manager, Mid-Level

  • Discord (San Francisco, CA): Senior Data Engineer, Senior-Level

šŸ“’FINAL NOTE

FEEDBACK

How would you rate today’s email?

It helps us improve the content for you!

Login or Subscribe to participate in polls.

ā¤ļøTAIP Review of The Day

ā€œBroad, yet detailed! Very interesting stuff.ā€

-Carolynn (1ļøāƒ£ šŸ‘Nailed it!)
REFER & EARN

šŸŽ‰Your Friends Learn, You Earn!

You currently have 0 referrals, only 1 away from receiving šŸŽ“3 Simple Steps to Turn ChatGPT Into an Instant Expert.

Reply

or to participate.