• The Prohuman
  • Posts
  • Microsoft Finds a Way to Cut AI Coding Costs

Microsoft Finds a Way to Cut AI Coding Costs

Plus: Audible Wants AI to Recommend Your Next Book

In partnership with

Hello, Prohuman

Today, we will talk about these stories:

  • Microsoft's Fix for Wasteful AI Coding Agents

  • Audible Adds AI to Help Listeners Choose Books

  • Google Finds Hackers Automating Exploits With AI

Frontier models perform better in Attio’s agent harness.

We gave GPT-6 Astra the same job twice: analyze 100 sales calls and draft the follow-ups, once in Ask Attio and once in ChatGPT.

Ask Attio finished in 4 minutes 49 seconds. ChatGPT took 26 minutes 11 seconds, about 5.4 times as long.

The most ambitious GTM work requires more than frontier intelligence. It requires architecture that can put that intelligence to work at scale.

Microsoft Wants Coding Agents to Stop Repeating Work

image credits: Microsoft

AI coding agents can waste money rediscovering the same information every time they work on a familiar codebase.

That gets expensive.

Microsoft Research and UC Berkeley developed TokenEconomizer, an open-source plugin that remembers useful details from previous coding sessions, removes unnecessary tools, and helps agents recover when they repeatedly fail.

Across 277 bug-fixing tasks, it reduced costs by 18.4% to 30.9% while keeping task success rates broadly stable.

I think the practical value here is significant because developers often return to the same repositories, yet their agents keep spending tokens on work they've already completed. Even remembering which test command to run in a terminal window can save repeated effort.

There's a catch. The system needs several sessions to learn, so early tasks can actually cost more. If these savings hold across larger projects, teams could reduce their AI coding bills without switching models or sacrificing reliability.

The bigger question is how much wasted spending comes from agents simply forgetting what they've already learned.

Audible Wants AI to Help You Pick Your Next Audiobook

image credits: Audible

Audible is introducing AI-generated audio previews that explain what a book is about and what other listeners liked about it.

You can listen before deciding.

The new feature combines book summaries, public reviews, and recorded commentary from authors or Audible editors, giving users more information without requiring them to scroll through reviews.

It launches first for select English-language titles in the U.S., as Spotify expands its audiobook offering to more than 180 markets. I think this makes sense for people browsing audiobooks while commuting or walking, when listening to a short explanation is easier than reading several reviews.

But there's a concern. If AI mainly highlights positive reactions, listeners could end up hearing recommendations that leave out legitimate criticism. With Spotify competing for audiobook listeners, Audible has good reason to make finding the next title easier.

The question is whether these AI previews will help people make better choices or simply encourage them to buy more books.

Hackers Are Using AI to Turn Software Patches Into Exploits

image credits: Google blog.

Hackers are using AI agents to examine software patches and figure out how to exploit the vulnerabilities those patches were meant to fix.

The process is getting faster.

Google's Threat Intelligence Group found attackers using autonomous frameworks like ARTEX, CyberStrikeAI, and PentAGI to automate parts of cyberattacks, including vulnerability analysis and exploit development.

The activity includes attacks involving PaperCut NG/MF vulnerabilities disclosed in September 2026.

What concerns me is how little manual effort this could eventually require, especially when publicly available patches give attackers a clear starting point for finding vulnerable systems. A newly released security patch can reveal enough technical detail for AI tools to help attackers develop exploits before many organizations have updated their software.

That matters for defenders. Security teams may need to shorten patching delays and respond faster to newly disclosed vulnerabilities as these tools become more accessible. How much time will organizations realistically have to patch before automated attacks begin?

Prohuman team

Covers emerging technology, AI models, and the people building the next layer of the internet.

Founder

Writes about how new interfaces, reasoning models, and automation are reshaping human work.

Founder

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.