AlphaSignal

Hey {{FIRSTNAME}},

Two massive things dropped today, and we think they rhyme.

MCP just went stateless, making it a real production-grade protocol. AI agents can now scale like normal web services. Meanwhile, Claude Mythos cracked post-quantum encryption that stumped human experts for two years, autonomously, in 60 hours.

The infrastructure is maturing fast. So is the capability. Both at once is worth paying attention to.

In Partnership with

Today's Author

Lior Alexander. Founder of AlphaSignal and former ML Engineer. Previously built ML systems at Iguazio, Guesty, Enphase, Mila.

Summary

Read time: 6 min 35 sec

Top News

▸ MCP goes stateless in its biggest update yet, unlocking serverless deployment

Mistral

▸ Agentic Coding That Meets You Where You Work

Top News

▸ Anthropic's Claude autonomously cracks encryption schemes in under a week

Datadog

▸ Build Trustworthy AI Agents, Faster

Top News

▸ Cursor launches ₹649/month plan with cloud agents and Grok 4.5 for India

Signals

1. Claude builds a full AAA-style space game in one 24-hour coding session

2. Tiger Data: TimescaleDB extends Postgres so queries run in milliseconds, not minutes

3. Microsoft open-sources a graph-based RAG tool to query private text data

4. Study finds AI-written books now grab 20% of Amazon sales, hurting human authors

5. LiquidAI releases multilingual encoder models in 230M and 350M sizes

6. Open-source text-to-CAD tool hits 10,000 stars, outputs STEP, STL, and gcode

Top News

MCP goes stateless in its biggest update yet, unlocking serverless deployment

9,630 Likes

MCP just dropped its biggest update ever. Here is what actually changed and why you should care.

The old MCP required a persistent session between client and server, meaning each user was locked to one specific server instance. Load balancing was basically impossible, and serverless deployments were off the table.

Now MCP is stateless. Every request carries everything it needs, so any server instance can handle any request. You can now deploy behind a standard load balancer, run on serverless platforms like Cloudflare Workers, or scale horizontally without a shared session store.

The big new things you get:

Extensions framework: MCP Apps let servers render UI inside a sandboxed iframe, and Tasks handles long-running async operations
Enterprise Managed Auth: plug in your existing identity provider to control server access centrally
Formal deprecation policy: future changes will not silently break your code

Updated SDKs for Python, TypeScript, Go, and C# are already available. Older clients still interoperate via automatic fallback.

forward →

Presented by Mistral

Agentic Coding That Meets You Where You Work

You describe the task. Vibe sees it through.

Vibe by Mistral is a coding agent that writes, tests, and deploys code autonomously with full codebase context.

Launch remote agents from a dedicated web surface and see sessions through to a pull request
Run multiple sessions in parallel — they persist while your machine is off
Integrate with GitHub, GitLab, Jira, and proprietary codebases
Inspect diffs and manage sensitive actions in an isolated sandbox
Use in your terminal or IDE — ACP-compatible

GET STARTED

partner with us →

Top News

Anthropic's Claude autonomously cracks encryption schemes in under a week

8,248 Likes

Anthropic's unreleased Claude Mythos Preview just did something that years of expert human review could not: it found real mathematical cracks in two encryption systems used to keep data secure online.

Here's what it cracked:

HAWK is a next-gen digital signature scheme built to survive future quantum computers. It passed two years of expert review. Mythos found a hidden mathematical shortcut in 60 hours, cutting its effective security strength in half.
AES is the encryption standard securing most internet traffic. Mythos found a way to speed up attacks on a research version of it by 200-800x, autonomously inventing a technique it named the "Mobius Bridge."

Neither breaks anything in production today. HAWK is not deployed yet, and the AES attack only hits a weaker research variant, not the real thing.

The bigger story: Mythos did most of this work on its own, with minimal human input. Each result cost roughly $100,000 in API usage. This opens up AI as a serious tool for stress-testing encryption before it gets deployed at scale.

forward →

Presented by Datadog

Build Trustworthy AI Agents, Faster

Small changes to an LLM can shift response quality, latency, cost, and trust.

As agents move from prototype to production, your team needs to see how they actually behave.

Datadog Agent Observability helps you monitor, evaluate, and control agent performance at scale.

GET THE BRIEF

partner with us →

Top News

Cursor launches ₹649/month plan with cloud agents and Grok 4.5 for India

7,147 Likes

Cursor just dropped Cursor Start, a new plan built specifically for India at ₹649/month (roughly $7). That is less than half the price of a Netflix subscription, and way below the $20 Pro plan.

Why does this exist? Simple. Cursor's India user base tripled to over 3 million users, making it their third largest market, with more agent requests per person than anywhere else on earth. They listened.

Here is what you get:

Grok 4.5 and Composer access, enough to build with AI agents every single day
Always-on cloud agents that build, test, and open pull requests while you sleep
Cursor for iOS so you can launch and steer those agents from your phone
Plugins, MCP servers, hooks, and skills to wire Cursor into your own workflow
Pay in INR via UPI, credit, or debit card

The catch? It does not include frontier models from OpenAI or Anthropic, or advanced features like Bugbot, Auto Mode, and Automations. That stays Pro-only. Existing free users can upgrade from the dashboard, new users pick it during signup.

forward →

Signals

Claude builds a full AAA-style space game in one 24-hour coding session

3,227 Likes

TimescaleDB extends Postgres so queries run in milliseconds, not minutes

Presented by Tiger Data

Microsoft open-sources a graph-based RAG tool to query private text data

34,958 Stars

Study finds AI-written books now grab 20% of Amazon sales, hurting human authors

187 Likes

LiquidAI releases multilingual encoder models in 230M and 350M sizes

5,327 Downloads

Open-source text-to-CAD tool hits 10,000 stars, outputs STEP, STL, and gcode

2,821 Likes

At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 300,000 developers stay informed and ahead. We're passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.

Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We're based in the United States. By using our site, you agree to be governed by U.S. laws.

Looking to promote your company, product, service, or event to 300,000+ AI developers?

WORK WITH US

How was today's email?

Awesome Decent Not Great

unsubscribe_me(): return True

{"AlphaSignal": "214 Barton Springs Rd, Austin, USA"}

Claude autonomously broke post-quantum encryption experts missed for 2