SMW.AI

Bob McGrew explains that AI progress relies on pre-training, post-training, and reasoning.

A short history of LLLM milestones from Davinci to Gemini Flash

Someone has an axe to grind through compiling existing content into a single page.

Two-hour interview with Tim Phillips, former Deputy Director of AARO, discussing his tenure at AARO and the state of government UAP investigations.

I wanted a fun dither/CRT effect for the site, but after searching GitHub, realized there wasn't an easy-to-use component to pull from.

A beautiful deep dive into javascript-based nostalgic visual effects.

Outside of code generation, my top daily use case for LLMs is making sense of information.

Relive the best of PC games with engines ported to the Mac.

Downloaded Xcode 2025 and found that Swift Assist is still nowhere to be found.

The internet laughed at Apple this weekend due to their terrible progress in AI and the irony of publishing research taking on today's reasoning models.

Never thought Notepad would get updates.

WWDC kicks off next Monday.

Google continues to dominate benchmarks with its frontier Gemini 2.5 Pro 0605 model.

Somewhere around the release of GPT-4 Turbo the default response format changed.

Anthropic is back — finally — with Sonnet and Opus 4.

After months of using o3-mini and o3 via deep research, the full o3 is here. o3 is the first model that feels more intelligent than me when I use it.

September 2024 — the FUD surrounding model scaling and performance returns filled the X/Twitter zeitgeist.

10 months after announcing of GPT-4o's omni model capabilities, OpenAI publicly released native image generation capabilities for GPT-4o.

In the wake of Grok 3, this feels like a step backward.

Grok 3 is finally here, and answers the question of the year: are scaling laws over?

In 48 hours, my opinion of Deep Research went from a disappointing, waste of money to something truly unique.

Shot on Kodak Gold 200 film

The Information and others leaked internal strategy realignment at OpenAI and Anthropic due to alleged diminishing model returns.

This post is now archived.

tl;dr Solving a problem you genuinely love, using today's free tools to build quickly, and keeping relentless focus makes turning a project into a scalable company uniquely achievable right now.

When you are so annoyed by a movie that you spend the next 24 hours in Final Cut Pro, recutting the film with 10 mins trimmed, adding a new score, and focusing the internal timeline.

This original article was published in early 2022.

It's Friday, I don't know how to make NFTs.

Late to the game, but The War of Art is one of my favorite books this year.

In a conversation with startup founders, I was asked my perspective on what makes someone 10x more effective.