Greetings and salutations!

Welcome to Pedal Assist Coding, a sporadic newsletter covering ethics and DevSec in the context of AI and coding agents, with a bit of a bent towards Kotlin Multiplatform.

New Newsletter, Who Dis?

I am Mark Murphy. Since 2008, I have operated as CommonsWare, originally serving as a freelance developer advocate for Android. I wrote books and Stack Overflow answers, back when those mediums were more relevant. I delivered professional training, back when that involved live people in rooms. I presented at conferences — which is still a thing, surprisingly enough. I contributed in small ways to open source projects and did all the other things that developer advocates do.

And, in the new world order, I need to re-tool a bit. This newsletter is helping to document that re-tooling to a world of coding agents.

Yeah, But AI Sucks!

I hear you. I had that mindset for quite some time, since the introduction of ChatGPT and Stable Diffusion and seeing all of the absolute crap that those things unleashed. Circumstances pushed me to reconsider my priors. So, I am delving into the world of coding agents, LLMs, and related tooling.

However, a large chunk of the world of software development is as high as a kite on AI. The fact that "tokenmaxxing" is somehow a valid sequence of characters is appalling. Coding agents as a development tool — think the next evolution of the IDE — is reasonable. Lots of people are being rather unreasonable in this space.

So, consider this newsletter to be a push-back both against those who feel that AI has no place in society (sorry, that genie is not going back in the bottle) and that you should "AI all the things".

What To Expect

I will publish this newsletter semi-regularly. Whereas my Jetpack Compose/Compose Multiplatform newsletter comes out pretty consistently at 9am US Eastern on Tuesdays, Pedal Assist Coding will be a bit more organic. I will be aiming for a post a week or thereabouts, perhaps bumping to two per week if I start a Thursdays with Koog series to talk about embedding LLMs into Kotlin apps.

Newsletter issues will touch upon:

• Coding agent-related tooling: tools I created, tools that I use, tools that I considered and rejected

• How to reduce reliance on frontier models and move more coding agent work to locally-hosted models, where you have greater control and somewhat fewer ethical issues

• How to avoid turning your development environment into a no-holds-barred OpenClaw type of environment, where if things run amok, your life will suck

And my next issue will explain a bit more about the name of the newsletter, explaining what I mean by the phrase "pedal assist coding".

Also, expect bad jokes. Lots of bad jokes.

What Not to Expect

I am not anticipating doing third-party advertising here. If I do, it would likely be along the lines of the ads that Simon Willison runs today: small uninterrupting-but-convenient banners with textual pitches and simple links. I do expect to point to things that I can do to help you, though.

I definitely am not expect to start charging for this newsletter. I may start up a subscription system in the future, reminiscent of my old Warescription, but this newsletter will not be part of that.

And, I do not intend to depict pelicans riding bicycles, though I may mention models that can generate such images.

(note: no actual pelicans or bicycles were harmed in the creation of this newsletter issue)

(also: did I mention that this newsletter will contain bad jokes?)

(and parentheses?)

Thanks!

If you subscribe, or get the newsletter through a feed or a social network or something, thanks for reading!

And now, if you will excuse me, I will get back to building my own coding agent harness... itself a subject for future issues.