Fermat's Last Theorem Proved in Lean 4: Sept 4, 2026

Read the HTML overview from GitHub - rendered via https://htmlpreview.github.io. 

Overview · FLT in Lean 4

A rival attempt to achieve the same outcome has ended in admitted defeat.

Almost 30K theorems were traversed on the path to the proof. The theorem statement is as follows:

theorem fermat_last_theorem (n : ℕ) (hn : 3 ≤ n) (a b c : ℕ) (ha : 0 < a) (hb : 0 < b) (hc : 0 < c) : a ^ n + b ^ n ≠ c ^ n

GitHub only shows raw HTML source (it does not render these in case of XSS and malicious scripts). HTML is always displayed as raw text.

Codex Terra Light 5.6 on the ChatGPT Application

A lightweight coding model (Codex Terra Light 5.6) is available in the ChatGPT app. It is ok - and can do very simple refactoring when you know the approach you want to take.

AI Orchestrators

LangChain, CrewAI, AutoGen, ADK and LangGraph are examples of tools used for agent orchestration.

Copliot Studio and Cost Analysis

Copilot Studio is a Microsoft product for building and managing AI agents. 

It has a corresponding licensing guide enabling you to estimate costs of building and running agents with it. At time of writing the licensing guide has been updated in July 2026.

Classical Reinforcement Learning and the Reward Signal

As a prelim to understanding RLHF, it pays to understand classical reinforcement learning, which has been studied under many different guises within computer science and more traditional engineering.

This is concerned with how an agent should take actions in a dynamic environment in order to maximise a reward signal.

Document Intelligence Infrastructure

Article covers GraphRAG and an agentic approach as a path to document intelligence,

GitHub's Agentic Workflows in the Crosshairs of GitLost

Noma Security has publicised GitLost, a technique to get GitHub Agentic Workflows to spill private repository data.

The exploit is aimed at organizations with both public and private GitHub repositories, and relies on prompt injection.

Specifically, it uses indirect prompt injection  where malicious instructions are injected into seemingly regular requests.

GitHub's Agentic Workflows introduce LLMs into GitHub Actions, which is how the exploit is enabled.

Navigating OpenClaw Architecture

The best place to start is by reading architecture.md in the docs/concepts subdirectory.

Then move on to the other concepts (not all equally weighted in importance) than read the source code.

The Powerful System.Environment

System.Environment is an all-powerful class.  

It allows interaction with the current environment and platform. It is something to revere when developing intelligent agents that need to navigate the user's environment.

System.Environment.SpecialFolder is a well-known enum.  

It contains the CSIDLs (these are not .NET "things", they are Windows "things"). 

CSIDL = Constant Special Item ID List.  Here are some of the popular values.

Desktop 0  (logical desktop rather than physical file system location)

ApplicationData 26

LocalApplicationData 28

System 37

ProgamFiles 38

A full rundown of CSIDLs is here.

OpenCode and GLM Models

OpenCode is an open source AI coding agent. It connects to free models and also allows connection to commercial models. GLM models are supported.

Amazon Bedrock

Amazon Bedrock is the AWS fully managed platform for building generative AI services. It is basically (like LangChain) a unified API to call into various models and provides managed capabilities like retrieval-augmented generation (RAG), vector embeddings, guardrails and agents/agent orchestration.

Windows Baseline Security

Windows Baseline Security is something new for the Age of AI Agents.

The AI Agent Prevention Society

One of the obstacles to AI agents becoming all powerful is the technology infrastructure that actively seeks to obstruct agents.

This technology has various names, one being WAF, or web application firewall.  

The purpose of a WAF is to stop web applications from common attacks. 

Recall the OWASP Top Ten Risks? Some of the motivation behind these protections are good - for example, guarding against bot-driven DDoS attacks.

There are various vendors and products in this space; such as Imperva WAF, Cloudflare, Akamai Site Defender, AWS WAF, Azure Web Application Firewall and Google Cloud Armor.

These solutions are all effective at detecting and blocking agents, headless browsers, scrapers, credential stuffing bots (where leaked usernames and passwords are used across numerous websites to exploit potential duplication) and automated (even if legitimate) login attempts.

Action taken could range from limiting requests based on IP, presenting CAPTCHA challenges and blocking certain geographies.  However, these actions may prevent legitimate bots from using required services, or performing agentic actions on behalf of legitimate users.

Personal AI Assistants

Personal AI assistants are here, but security controls are askew.  

OpenClaw is an example of such an assistant.  It acts as a bridge between messaging devices (such as WhatsApp, Telegram, iMessage) and AI agents.

You will find OpenClaw "advertised" on openclaw.im (the im suffix indicates the country code top-level domain (ccTLD) for the Isle of Man. You don't need to be in the Isle of Man to have this domain, and it had become quite popular with instant messaging software with registrations from Meebo and Yahoo! amongst others (Meebo was acquired by Google in June 2012 at which point it had 100m users).

Some of its taglines include "Your Code, Your AI, Your Rules". Its GitHub location is here.

Ollama integrates with OpenClaw. In this context, Ollama acts as a local LLM provider.

Microsoft Launch Fara-7B: A CUA (Computer Use Agent) in SLM Form

And here we have it. Ready for action on Hugging Face, Sir.

Sandboxing and monitoring are recommended. The agent itself is a wrapper around Playwright.

Microsoft Semantic Kernel

Microsoft Semantic Kernel is a "lighweight, open-source development kit" to build AI agents and integrate models into C#,  Python and Java code.

When you load up SK into a fresh Visual Studio Code (no extensions) it will prompt to install recommended extensions. These will include:

• ESLint - integrates ESLint JavaScript into VS Code (for static analysis)
  
• Prettier - integrates Prettier, the opinionated code formatter (for JavaScript, TypeScript and other webby stuff)
• Azure Functions - to quickly manage serverless apps directly from VS Code
• vscode-pdf - to display pdf files in VSCode (required to open PDF code maps for .NET and Python)

Cost Effective Deployment of Language Models

Cost effective deployment of language models (explicit financial as well as implicit environmental cost) is partly responsible for triggering the interest in small language models (SLMs) as alternatives for specific applications. 

Nvidia Research have a great paper on this entitled "Small Language Models are the Future of Agentic AI" with the recommendation that more routine tasks (non reasoning tasks) move from LLMs to SLMs. 

Fine tuning these SLMs for specific tasks can also enhance the effectiveness of deployed models.

The Model Context Protocol (Merci, Anthropique)

The MCP or Model Context Protocol was introduced by Anthropic as a way of sharing data with LLMs, or put differently, connecting LLMs to wider data sources. 

Anthropic has dubbed it a "universal translator" / "USB-C port" for AI applications. Your chat application can now talk (in a standard way) to your database (as an example).

MCP is also highly relevant for those developing AI agents. A standard protocol makes integration easier. 

MCP was mentioned by Sundar Pichai in Google I/Os 2025 keynote.