Joshua Wood

Microblog

This case study with Nate at ClickFunnels is a great example of how to validate production changes before they reach customers.

Today in reasons to lock down your home network and sandboxes: imagine when your 10yo gets their hands on an LLM

Do you think Trump ever bops Stephen Miller on the head and says stop hissing in my ear

Now that we don’t write code I’m worried my Ruby is getting Rusty

Deploying kernel update to dev box that has 5 PRs in flight, it’s fine

Was there an Apple thing today? I haven’t paid attention to one of those in years. That’s kinda wild.

Tomorrow I’m rebooting the Honeybadger newsletter: short emails about how we build and run a monitoring service with ~3 devs.

First issue: shipping risky changes without breaking production.

Wanna follow along?

Did a bunch of work on the NAS last night. Updated media profiles, organized files, etc. Freed up about 5 TB. Also rediscovered my lossless music collection. Maybe it’s time to start collecting again.

Went to see Wildwood (amazing), naturally ran into the Unipiper afterwards

Unipiper dressed as darth vader in a kilt holding bagpipes while peddling a unicycle

calling it a “brilliant but playful docudrama” and comparing it favorably to The Social Network (2010)

ok please fuck right off (new movie about Sam Altman/OpenAI)

Blessed are the tastemakers: for they shall be called the children of Claude.

honeybadger in zed and herdr on my mac running in a nix shell inside a remote nix host over LTE via tailscale, what a time to be alive

also with rootless docker + local port forwarding for worktrees

Hopefully I’m approaching the plateau of productivity with LLMs, and not the precipice of pathology

This is just an idea but what if you don’t fund an X-Files unit if you’re trying to hide government secrets

This morning while working on two other projects, I:

  1. Migrated my blog from Jekyll to Astro

  2. Added support for standard.site (atproto) documents/publishing and bluesky previews

  3. Rewrote a local ruby API that I originally built by hand for an arcane protocol (MetaWeblog+MoveableType) that allows me to manage the Astro content in my GitHub repo via MarsEdit. Previously it just ran locally and worked directly with the file system. Now it’s TypeScript in Cloudflare workers and works anywhere.

Also finished designing a landing page, built an install wizard for Honeybadger agent skills, and improved the UX of my lil native/SwiftUI side project.

Not a bad Saturday, time to go touch grass. 😅

Taking time off this summer was so nice. The industry has changed a lot recently. Too much, too fast, tbh. It was good to step back to gain some perspective.

Things I still believe about software:

  • good software takes effort
  • slower is often better
  • the hard work is human
  • it should make people’s lives better

At the same time, software is changing, and it can change for the better. It’s really up to us. If we can now do even more with less, we should aim higher.

There are plenty of new problems to solve, and I’m excited to dive back in. Trying to make the process of developing software a little better every day.

Haven’t been posting much lately because I’ve been vibe coding a cross-posting app. Preventing me from posting is its #1 feature.

Personal data sovereignty is becoming my number one personal computing concern lately, heh. Like I’m still fine using apps/SaaS, but if they lock up my data, they can fuck right off.

Nicolas Cage standing in front of the Lincoln Memorial Reflecting Pool, which is reflecting the Washington Monument

Watching National Treasure in honor of America’s 250th birthday. 🇺🇸 (source)

Watching National Treasure in honor of America’s 250th birthday. 🇺🇸

Spending the first couple days catching up on life things I was neglecting… switched insurance providers, got a bid to fix the roof, did some organizing.

I’m taking some time off from work to recharge this summer, very much appreciate my team for making this possible. (source)

I’m taking some time off from work to recharge this summer, very much appreciate my team for making this possible.

No one told me InfoWars is doing guests, I could be the second member of my family to be on InfoWars.

Good news, folks. We have closed all the gaps. There are no more gaps left to close. Might as well take the day off. Happy Friday!

It’s always funny to me when our VC/PE-backed competitors target us directly in their marketing. Like we’re five people, bro. You should probably be concerned that you think about us at all.

“The reason I’ll never forget Jon is he wrote his name on my shoe.“ —8yo on last day of school

Is Honeybadger building another agent?

We’ve been thinking a lot about AI and the future of application monitoring at Honeybadger. The past few years have been chaotic, fascinating, and everything in between.

Many in the industry claim that the future is agentic everything, but no one knows what the future holds. We’re still figuring this out just like you.

One thing is clear: AI is extremely useful for monitoring and debugging.

But our goal is the same as it’s always been — to cut through the noise and continue delivering the human-centric product and support that we have for 14 years.

So no, we’re not building another agent. Instead, we’re providing structured, queryable, signal-rich data that works with your existing LLMs and automation so that you (and your agents) can get to the root cause faster.

Does this resonate with you?

PS: We reject the idea that AI will replace developers. If anything, skilled developers are more important than ever. We’ll always be on the side of developers.

On day 2 of decaf and feel fine. I tapered to half caff in the morning for three or four days first. Headache was actually not bad. Definitely feels calmer so far.

Thinking about quitting coffee/caffeine. Has anyone done this? (source)

Software leaders on LinkedIn are starting to talk about the growing quality crisis in agentic engineering and they’re like, “It’s not just a crisis, it’s a five-alarm bill that’s coming due. The gap between…”

If I see one more slop post capitalizing on the growing quality crisis in agentic engineering, I’m going to lose my mind. This shit is affecting real people. The least you can do is take the time to say it yourself.

Thinking about quitting coffee/caffeine. Has anyone done this?

I wrote a little Ruby script today by hand. It was fun and rewarding, I recommend it.

I wish I could get my kids to ELI5. Explain it like I’m five, not explain it like you’re five!

I am currently editing our company blog after our last editor left, and realized some AI content was slipping through. So now we have a policy for authors. A few thoughts:

As a reader, slop is cringe. When I have the feeling that something is AI-generated, I nope out immediately. Everyone I’ve asked does too. (If I know you and you’re doing this, oof.)

If it’s not immediately obvious but I realize you’ve tricked me, I am now physically mad at you. You wasted my time. I will go out of my way to avoid your content in the future.

We’ve also found that when authors start with AI in their writing process, it makes it nearly impossible to edit or provide review feedback. The logic of the piece is just a disjointed mess. The only feedback I can give is to throw it away and start over.

For these reasons, we do not allow our authors to use generative AI for content. We do allow very limited use for research and codegen, but that’s it.

Honestly, if you consider yourself a writer and you use AI to generate your words, you are not a writer.