Gavin Morgan

Hey, I would love to offer some contributions to this. Is it extensible? Or, are you open to pull-requests?

Exciting! Any major architectural changes?

@jmcodes.bsky.social Jesus, are you still using lootbox, or have you migrated entirely to skills, or something?

I like the Unixy flavor of lootbox, so I've started to explore it for my agents, but I want to hear from the man himself!

I think a fundamental issue with modern metaphysics is that we don't consider our mind a home to be comfortably furnished.

A well-appointed mind is a treasure unto itself, to say nothing of its value in concert with others'.

An infographic illustration titled "The Princess and the Wishing Well: An Allegory of Hexagonal Architecture." The image uses a colorful storybook scene to map software architecture concepts to fairytale elements: ​Center (The Core): A princess sits on a throne inside a stone gazebo. A text box reads: "The Princess is the Core Business Logic. She embodies all business rules and makes decisions, completely isolated from outside concerns. The Core is Technology-Agnostic. The princess is 'blissfully ignorant' of how her requests are fulfilled or where commands come from." ​Top Left (Ports): A scroll banner hanging from a tree reads: "The Royal Language is the 'Port'. The princess defines the strict interfaces (the 'what') for all communication with her." ​Left Side (Inputs): A crowd of villagers hands scrolls to royal advisors. Text reads: "Driving Adapters: The Court Advisors. They take requests from the outside world (users, texts) and translate them for the princess." A stream of magic flows from the advisors to the princess. ​Right Side (Outputs/Infrastructure): A frog wearing a crown and holding a wand sits on the edge of a stone well. Text reads: "Driven Adapters: The Magical Frog. The frog takes the princess's wishes (the 'how') and adapts them to the wishing well (database)." The side of the well is cut away to reveal gears and machinery inside, labeled "Database."

Really getting a kick out of generating technical infographics w/ NotebookLM.

Is there really an expert level of Python? Am I missing something by not having read Fluent Python before now? On this flight to my Thanksgiving destination, I hope to find out.

(Tim Peters, legendary core developer and author of "The Zen of Python")

From the forward of Fluent Python. Python was the 2nd programming language I learned (not counting Matlab, which I never really picked up), and I've used it on-and-off throughout my career, but never really felt expert in it.

"Here’s the plan: when someone uses a feature you don’t understand, simply shoot them. This is easier than learning something new, and before too long the only living coders will be writing in an easily understood, tiny subset of Python 0.9.6 <wink>."

🫱🏼‍🫲🏽 Great work necessitates the development of teams’ social dimensions, prior to and alongside their task dimensions. ⚖️ Striking that balance is part of the challenge of engaging in a profession. 👔💼

Update your jargon file:

There are 3 hard problems in software engineering: naming things, off-by-one errors, cache invalidation, and polymorphic lists.

A maxim of the internet age: to get answers don't ask questions. Instead, make an incorrect assertions.

How to do this without gaining a reputation as a bold ignoramus?

We're dependent on colleagues' proclivity for Fundamental Attribution Errors, but we can reciprocate by our own graciousness.

For heaven's sakes, Garmin—just let me fetch my step count from a raspberry pi, already!!

Can anyone recommend a fitness wearable with reasonably hackable APIs? I'm trying to create information radiators for my home.

Looking forward to images from ESA’s Jupiter Icy Moons Explorer of interstellar body 3I/ATLAS in November!

MCP is a way of intercepting control flow from LLMs and passing it back to expert systems.

932: Vibe Coding’s Huge Problem

"We're in for a rough couple years of these vibe-coded apps having major security issues."

Objects of indiscernible shape obscured by frosted glass

The fundamental nature of AI as a product may be its ability to stingily titrate access to what Lux's Josh Wolfe calls "longitudinal repositories of siloed information."

The essential benefit of this titration is certainly not latency. It may be intelligence—greater penetration—or obfuscation.

I didn't see it, but other similar products.

Is Genie 3 being demonstrated at SIGGRAPH?

I just happened to be in Vancouver this week. 😬

Is there any web technology stuff I should keep a lookout for?

Oops, I wound up at #SIGGRAPH! What's an absolute must-see? Or, I guess, what are you at?

Finally, another open model from OpenAI; I'm super excited to try out openai/gpt-oss-120b... just as soon as I get home from hanging out with the family, this week.

How can I keep four machines' configuration and utilities in sync? Still Chef?

Where is it even ergonomic to host an Elixir app? I'm of the mindset that Cloudflare edge workers are a nifty place to modularize "actor-like" business logic into, but what sort of PaaS maximally exploits the BEAM vm?

Comparison of nominal sizes of primary mirrors of notable optical reflecting telescopes, and a few other objects. Dotted lines show sizes of round mirrors that would have had equivalent light-gathering ability. The telescopes shown on this comparison chart are listed below, ordered in each sub-section by (effective) mirror/lens area, low to high, and then by actual/planned first light date, old to new. The "present-day" status is given as of the beginning of 2024. See also List of largest optical reflecting telescopes. Largest refractors (for comparison): 1) Yerkes Observatory's 40-inch (1.02 m) refractor, 1893 (largest refractor consistently used for scientific observations) 2) Great Paris Exhibition Telescope, 49 inches (1.24 m), 1900 (largest refractor ever built; had practically no scientific usage) Ground-based reflectors: 3) Hooker Telescope, 100 inches (2.54 m), 1917; world's largest telescope from 1917 to 1949 4) Multiple Mirror Telescope, 186 inches (4.72 m) effective, 1979–1998; 6.5 m, from 1998 5) LAMOST (Large Sky Area Multi-Object Fiber Spectroscopic Telescope), 4.9 m effective at best, 2009 6) Hale Telescope, 200 inches (5.1 m), 1949; world's largest telescope from 1949 to 1975 7) BTA-6, 6 m, 1975; world's largest telescope from 1975 to 1990 (when it was surpassed by the partially-completed Keck I telescope) 8) Large Zenith Telescope, 6 m, 2003; largest liquid-mirror telescope ever built; decommissioned in 2019 9) Magellan Telescopes, two 6.5‑m individual telescopes, 2000 and 2002; 10) Vera C. Rubin Observatory (formerly Large Synoptic Survey Telescope), 6.68 m effective (8.4‑m mirror, but with a big hole in the middle), planned 2025 11) Gemini Observatory, 8.1 m, 1999 and 2001 12) Subaru Telescope, 8.2 m, 1999; largest monolithic (i.e. non-segmented) mirror in an optical telescope from 1999 to 2005 13) Southern African Large Telescope, 9.2 m effective, 2005 (largest optical telescope in the southern hemisphere) 14) Hobby–Eberly Telescope, 10 m effe…

@miquai.bsky.social I knew there were observatories all over, but I don't know why it never occurred to me that we never stop designing and building bigger, better ones!

Very exciting to consider, and the scope of the data brought in is absolutely mind-boggling.

What's it gonna take for me to finally break ground on my garden plot?

Remind me to experiment with using Djikstra's pathfinding for agent-to-agent transaction selection. Or, at least, some kind of "movement cost" (to approximate, among other things, distance) estimation for my next economic simulation.

Imperfect—but nonzero—information!

Running into this weird cognitive dissonance when it seems like people are using the term #agentic to mean "IDE". I know it sounds crazy! I must be missing a piece of the puzzle, or else is it just evidence that we're not yet seeing agentic tool adoption outside of software industry?

Frankly, however, it's extremely attractive and compelling the thought of an agent daemon living on my machine and planning changes in response to the context of my open files and keystrokes.

I think it's probably critical not to give agents right access to anything that isn't in Version Control.

It's very easy to imagine a local desktop agent getting the notion in its head that there's something extremely inefficient or disorganized about your file system and spontaneously deciding to reorganize all your files while you're not looking.