Ben Galbraith

Ben Galbraith

@octo.dad

Octodad and yet, surprisingly, younger than @almaer.com; lover of fine pies. Building, advising, parenting. Ex: Google, Walmart, Palm, Mozilla, Ajaxian, etc.

959 followers424 following140 posts

I feel the paradox of choice so much more intensely when I’m building hobby projects with agents. Because my own familiarity with a coding domain is no longer a useful razor at prioritizing, it’s very disorienting being able to do... anything!

I’ve come to dread the cognitive burden of new Apple hardware. I tell myself I don’t need to upgrade to the latest [phone / laptop / watch]. But then the daily struggle begins: “Wouldn’t it be nice to have that [new feature]?” “See that person? They have the new [y]; shouldn’t you have it?” 😭

🧵 AI coding agents — like Claude Code and Codex — are amazing, but they’re also unpredictable and hard to control. Today at Tessl we’re announcing a framework that helps keep agents on the rails and arms with the context they need to get things right. It has two parts:

View full thread

(2) a registry of specs that document thousands of open-source libraries, helping coding agents use third-party code with fewer hallucinations and research / try / fix loops.

We’re releasing these early — they’re very much a beta — and there’s a long journey ahead. We’re doing it in the open and excited to build this with a community of like-minded pioneers! Check it our launch post and let me know what you think: tessl.co/l5y

tessl.coAnnouncing Tessl's Products to Unlock the Power of AgentsTessl introduces AI native development where specs drive everything. Build reliable AI-powered software with autonomous maintenance, cross-platform generation, and self-optimizing performance. Request...

🧵 AI coding agents — like Claude Code and Codex — are amazing, but they’re also unpredictable and hard to control. Today at Tessl we’re announcing a framework that helps keep agents on the rails and arms with the context they need to get things right. It has two parts:

(1) a CLI + MCP server that provides tooling and context for a spec-driven development process: creating and editing specs (from prompts and existing code), keeping code and specs in sync, build tests, etc.

Not sure when Google Docs enabled right-clicking to copy embedded images to the clipboard, but I’m so grateful. All is forgiven, Google Docs. 🥳

I track things like the food I eat, shows I watch, how I spend my time generally, etc. Lately I switched from using custom apps for this (e.g., Sofa for media; MacroFactor for food) to just writing a free form log in a daily note and having my AI assistant analyze it. It’s so much better!

Finding it’s delightful to have put myself into two obscure categories, from an IT troubleshooting point of view: * Families with more than four children * Families who immigrate to another country but still want to have access to their original country’s digital assets Absolutely delightful.

Reviewing my car insurance. Says I have 32 years of driving experience. First reaction: “That’s wild, I wonder where they came up with that.” Second reaction: 🥺

In practice, the new iOS text effects are just another way for iOS to do something to my text that I didn’t want, and for me to spend extra cycles trying to undo it. Yay.

“Dad, why do so many company names start with ‘a‘?” Son, lemme tell you a story about these ancient things called “phone books”...

Can’t decide which blows my mind more: • the Apple ][‘s peak sales year was six years after its introduction • the Apple ][ was an active SKU for 16 years • the Apple ][‘s lifetime sales of 6 million units is routinely bested by a few *days* of iPhone sales

I used to find most headlines from The Onion laugh-out-loud funny, but lately the jokes just aren’t landing with me. Have I changed, or have they?

Weird to go from infinitely configurable custom feeds to “here’s a ton of threads that intersect in an undefined way between your interests and ours.”

Trying to grok the galaxy brain strategy Starbucks is employing that explains why their bakery display cases across the country are consistently gross.

There’s awful, and then there’s “incapable of using logic” and “posting videos of your lame workouts and then making excuses about them.” Two different leagues. JFK was a very rational, logical guy.

This post is unavailable.

This one hits me hard. I have fond memories of looking through dozens of back issues of Nat Geo when visiting my grandparents, and of looking forward to each month’s new issue. Still, times are changing, and I haven’t subscribed for decades.

Dare Obasanjo@carnage4life.bsky.social

National Geographic laying off its entire writing and podcast staff but still planning to keep publishing issues is bonkers. There’s no way they’ll be able to keep the quality up with freelancers & AI. This is a death spiral for an iconic magazine. Is there any future for journalism majors?

As I age, I wish more adventure games were *linear* and almost on rails. I don’t have much free time, and any time spent wandering around wondering what to do next and whether I’m on the right track winds up being stressful. Just keep feeding me puzzles on a straightforward narrative track.

5k lbs per sq inch is truly 🤯

Neil deGrasse Tyson@neildegrassetyson.com

Exploring space is easier than exploring the ocean bottom… Pressure difference between a space capsule and the vacuum of space: Less than 1 atmosphere. Pressure difference between a submersible at the Titanic and the surrounding water pressure: Nearly 400 atmospheres (5,600 pounds per square inch)

Disappointing article. No effort to contextualize the performance of Tesla self-driving / automation, vs. human driving. Just some absolute numbers while acknowledging that Tesla’s scale dwarfs any other party attempting similar automations.

washingtonpost.com17 fatalities, 736 crashes: The shocking toll of Tesla’s AutopilotCrashes have surged in the past four years, reflecting the hazards associated with increasing use of Tesla’s futuristic driver-assistance tech.

Loved this bit: “Mr. Gottlieb… often signaled changes with hieroglyphics in the margins: a wavy line for language too florid, ellipses or question marks advising a writer to ‘think harder and try again.’” https://www.nytimes.com/2023/06/14/books/robert-gottlieb-dead.html

nytimes.comRobert Gottlieb, Eminent Editor From le Carré to Clinton, Dies at 92At Simon & Schuster, Alfred A. Knopf and The New Yorker, he polished the work of a who’s who of mid-to-late 20th century writers.

“[some new startup] claims it is actually an AI play!” Shocking turn of events. 😉

Louis Gray@louisgray.com

The Instagram Founders’ New News App Is Actually an AI Play Artifact offers a personalized feed of news and can also translate clickbait headlines using AI. Kevin Systrom says that’s only the beginning. https://www.wired.com/story/plaintext-instagram-founders-artifact-news-app-ai/

This LLM stuff has me super uneasy about how often my brain generates words that I say, and then work out why I said them. Ideally it’s gen’d by some subsystem that actually has a goal aligned with something self-serving, but… how often is it just… pattern matching? 😬