AnalysisAISoftwareOutlookPersonal
The Golden Age of Personal Software
Code is getting cheaper by the month. Before full automation reshapes the industry, we're entering a window where custom software becomes accessible to everyone.
LMF Blog
Thoughts on AI, technology, and the things I build. For timeless concepts, see the Learn Guide.
Analysis
The Golden Age of Personal Software
Code is getting cheaper by the month. Before full automation reshapes the industry, we're entering a window where custom software becomes accessible to everyone.
AnalysisAISecurityAgentsNVIDIAOpen Source
NemoClaw: NVIDIA's Security Sandbox for AI Agents
NVIDIA open-sourced NemoClaw — a Landlock + seccomp sandbox for OpenClaw agents with declarative network policies and operator approval flows. Not a framework, but the missing security layer.
OpinionAIAgentsParadigmPersonal
How Free-Roaming Agents Are Changing the Meta
Give a language model a shell, persistent memory, and modular tools — then let it run. A look at how the agent-as-infrastructure pattern is changing how developers actually work.
AnalysisAIBenchmarksGemini
Gemini 3.1 Pro and the Downfall of Benchmarks
Gemini 3.1 Pro is an incredible model — but it exposes a deeper problem: benchmarks no longer tell us which AI is best. Welcome to the vibe era.
TutorialSecurityAIAgents
How to Secure Your Agents
A practical guide to securing AI agents — from prompt hardening to honey pot tools. Seven defense layers for every threat model.
OpinionAIVideoOutlook
Seedance 2.0: Stock Footage Is Dead
Seedance 2.0 doesn't just beat Sora, Veo, and Kling — it makes them look like last year's tech. The real story is what this means for video production, VFX jobs, and trust in visual media.
OpinionOutlookAI
When Intelligence Becomes Dirt Cheap
What happens to society when the cost of thinking approaches zero. White collar, blue collar, education, and the three possible futures.
OpinionAIOpen Source
The Real Heroes of Local AI
The engineers and projects doing the actual work that makes running LLMs on consumer hardware possible — and why they deserve more credit.
OpinionOutlookAI
Software Projects in the AI Age (2027–2028)
How software teams will restructure around AI agents: three project tiers, shifting skills, and the junior engineer problem.
News
February 2026: Models to Watch
Rumored releases, confirmed retirements, and what to keep an eye on this month in AI.
Analysis
Qwen3-Coder-Next: The Local Coding Agent That Actually Fits
80B parameters, 3B active. Qwen's new coding model is built for unified memory systems like Strix Halo, DGX Spark, and M4 Macs.
Analysis
Kimi K2.5: 1T Open-Source Model with Agent Swarms
Moonshot AI's 1 trillion parameter model with video-to-code, agent orchestration, and strong benchmark scores.
OpinionOutlookAI
Education 2.0: LLMs Teaching Children
Reimagining education with AI tutors, skill-based assessment, and teachers as social mentors.
TIL
Gemini 3 Flash: Perfect Long Context Scores
Gemini 3 Flash Preview achieves 100% on Fiction.LiveBench across all context lengths up to 192k tokens.
News
Hello World
Welcome to the LMF Blog. A space for thoughts on AI, technology, and the things I build.