Daily brief · Tuesday, July 28, 2026

Optimizing CPU Inference, Kernel Profiling, and Scaled Pipelines

Today's brief focuses on practical system optimization: running fast long-context encoder models on CPU, profiling eBPF kernel hooks with Linux perf, and scaling distributed inference pipelines.

14 min read7 storiesAI-assisted · editor-reviewed

0 / 7read

Category banner for the lead story · Agents & MCP

The brief in 30 seconds

Agents & MCPPass tool outputs by reference in custom agent harnesses to preserve prompt cache hits.
Tools & releasesInstall `i-have-adhd` to cut non-functional tokens from AI agent CLI responses.
Token & cost optimizationTreat all Claude share links as public web pages accessible to search crawlers.

Agents & MCP
2 min · NVIDIA Blog
NVIDIA Open-Sources NOOA: Python Object-Oriented Framework for AI Agents
NVIDIA Labs released NOOA, an open-source agent framework that defines agents as single Python classes with type annotations. By passing live object references instead of text dumps, NOOA cuts agent token consumption by half on SWE-bench Verified while reaching 82.2% accuracy.
Why it matters. Engineers can build cleaner agent architectures that eliminate context compaction and cut LLM API costs by 50%.
Full context
Tools & releases
2 min · GitHub
i-have-adhd Plugin Strips AI Assistant Fluff for Direct Code Action
The open-source plugin `i-have-adhd` enforces concise, action-first response formatting for Claude Code and OpenAI Codex. It removes polite filler phrasing, formats steps into numbered checklists, and provides specific time estimates.
Why it matters. Developers can save context tokens and reading time by suppressing chatbot preamble on every CLI session.
Full context
Token & cost optimization
2 min · TechCrunch
Public Claude Shared Links Search-Indexed: Audit Your Shared Artifacts Now
Google search indexed publicly shared Claude chats and Artifacts containing sensitive medical records, internal corporate docs, and source code via `site:claude.ai/share`. Developers should review their shared links in Claude settings immediately.
Why it matters. Developers using Claude Artifacts to share code snippets or specs must audit active share links to prevent accidental data leaks.
Full context
Models & research
2 min · Hacker News
Moonshot AI Kimi K3 2.8T Model Available on Telnyx Inference API
Telnyx launched hosting for Moonshot AI's flagship Kimi K3 model, featuring 2.8 trillion parameters, a 1M-token context window, and native vision. The model supports function calling and OpenAI-compatible API endpoints.
Why it matters. Developers can query a frontier 2.8T open-weights model with prompt caching and configurable reasoning effort via standard OpenAI client libraries.
Full context
Agents & MCP
2 min · MarkTechPost
Moonshot AI Open-Sources AgentENV: Firecracker Sandbox Infrastructure for AI Agents
Moonshot AI and kvcache-ai have open-sourced AgentENV under the MIT license, providing high-density Firecracker microVM sandboxes for agent execution. Featuring sub-50ms boot times and 16-way environment cloning, it includes an E2B-compatible API for drop-in orchestration.
Why it matters. You can self-host agent isolation infrastructure with E2B SDK compatibility without cloud sandbox costs.
Full context
Tools & releases
2 min · TechCrunch
Microsoft Announces MAI-Cyber-1-Flash Cybersecurity Model and Perception Agent System
Microsoft revealed MAI-Cyber-1-Flash, a 5B-active-parameter specialized model for code vulnerability detection, alongside Perception, an agentic security orchestration platform. Perception deploys coordinated Red, Blue, and Green agent teams inside the MDASH harness.
Why it matters. Automate code vulnerability scanning and remediation using multi-agent security teams.
Full context
Models & research
2 min · Simon Willison
Moonshot AI Releases 2.8T Parameter Kimi K3 Weights with Custom Licensing Terms
Moonshot AI has made the weights for its 2.8 trillion parameter Kimi K3 model publicly available on Hugging Face as a 1.56TB download. The release features a customized license requiring separate commercial agreements for large Model-as-a-Service providers and mandatory UI attribution for high-revenue apps.
Why it matters. Developers planning self-hosted or commercial API deployments of Kimi K3 must audit their revenue and distribution channels against Moonshot's custom licensing constraints.
Full context

One thing to try

Optimizing CPU Inference, Kernel Profiling, and Scaled Pipelines

The brief in 30 seconds

NVIDIA Open-Sources NOOA: Python Object-Oriented Framework for AI Agents

i-have-adhd Plugin Strips AI Assistant Fluff for Direct Code Action

Public Claude Shared Links Search-Indexed: Audit Your Shared Artifacts Now

Moonshot AI Kimi K3 2.8T Model Available on Telnyx Inference API

Moonshot AI Open-Sources AgentENV: Firecracker Sandbox Infrastructure for AI Agents

Microsoft Announces MAI-Cyber-1-Flash Cybersecurity Model and Perception Agent System

Moonshot AI Releases 2.8T Parameter Kimi K3 Weights with Custom Licensing Terms

Review NOOA agent harness design on GitHub.

Optimizing CPU Inference, Kernel Profiling, and Scaled Pipelines

The brief in 30 seconds

NVIDIA Open-Sources NOOA: Python Object-Oriented Framework for AI Agents

i-have-adhd Plugin Strips AI Assistant Fluff for Direct Code Action

Public Claude Shared Links Search-Indexed: Audit Your Shared Artifacts Now

Moonshot AI Kimi K3 2.8T Model Available on Telnyx Inference API

Moonshot AI Open-Sources AgentENV: Firecracker Sandbox Infrastructure for AI Agents

Microsoft Announces MAI-Cyber-1-Flash Cybersecurity Model and Perception Agent System

Moonshot AI Releases 2.8T Parameter Kimi K3 Weights with Custom Licensing Terms

Review NOOA agent harness design on GitHub.