Hello, there! đ
In light of the major, ongoing AWS outage⌠I hope youâre not having a terrible start to your work week. But if Downdetector is anywhere near accurate, those hopes are all but dashed.
Every time AWS goes down, Iâm reminded of the quip: âTurns out âthe cloudâ is just some building in Virginia.â
Ok, letâs get into this weekâs news.
Weâre joined by Deepak Singh from the Kiro team. Kiro is AWSâs attempt at building an AI coding environment to take you from prototype to production. It does that by bringing structure to your agentic workflow with spec-driven development. Their aim: the flow of AI coding, leveled up with mature engineering practices. đĽ VIDEO HERE đ
In an excellent piece designed to help engineering leaders and developers understand flow states and how to reclaim them, Csaba Okrona lays out exactly what Flow is:
Flow, as defined in the research, is âa psychological state of complete immersion and engagement in an activity.â For developers, itâs that magical zone where code seems to write itself, complex problems unravel naturally, and hours pass in what feels like minutes.
Then he enumerates the three major blockers to flow:
- Insufficient cognitive challenge
- Situational barriers
- Internal factors
And then shows you how to engineer your way back to Flow. Most of us canât get this done entirely on our own. In that case, forward this to your boss!
Letâs do this one âteaser trailerâ style:
Deep within the robotâs startup scripts, I discovered the smoking gun.
Inside the /etc/init.d/ directory, one script had been modified to prevent the main application from launching. This wasnât a glitch; it was an intentional commandâŚ
Someoneâor somethingâhad remotely issued a kill command.
Are you sufficiently teased?!
Ruby creator, Matz, shares some much-needed news for the Ruby community after the recent debacle (that we discussed in-depth on last Fridayâs show):
RubyGems and Bundler are essential official clients for rubygems.org and the Ruby ecosystem, bundled with the Ruby language for many years and functioning as part of the standard library.
Despite this crucial role, RubyGems and Bundler have historically been developed outside the Ruby organization on GitHub, unlike other major components of the Ruby ecosystem.
To provide the community with long-term stability and continuity, the Ruby core team, led by Matz, has decided to assume stewardship of these projects from Ruby Central. We will continue their development in close collaboration with Ruby Central and the broader community.
The wait is over. Zed for Windows is here!
For a long time, Windows devs have been asking the same question: Windows when? Well, now. The Zed team just dropped a native Windows build, built from the ground up with the same speed, multiplayer editing, and buttery-smooth experience that Mac and Linux users have been bragging about.
Whyâd they do it? Because great tools shouldnât care what OS youâre on. Zedâs mission has always been about fast, collaborative coding â and now, that magic extends to the worldâs largest community of developers.
So whether youâre pair-programming with your Mac-using teammate or just want an editor that feels instant on Windows, Zedâs ready.
Learn more and install Zed for Windows at zed.dev
Simon Willison is pretty excited about Anthropicâs recent announcement of Claude Skills â a simple Markdown system that teaches Claude how to do new things.
Claude Code is, with hindsight, poorly named. Itâs not purely a coding tool: itâs a tool for general computer automation. Anything you can achieve by typing commands into a computer is something that can now be automated by Claude Code. Itâs best described as a general agent. Skills make this a whole lot more obvious and explicit.
Simon goes on to explain how Skills compare to MCP, why he likes them better, and how Skill sharing might make this yearâs MCP rush âpedestrian by comparison.â In the end, itâs all about simplicity.
Skills are Markdown with a tiny bit of YAML metadata and some optional scripts in whatever you can make executable in the environment. They feel a lot closer to the spirit of LLMsâthrow in some text and let the model figure it out.
Luke Plant looks at everyoneâs favorite software-solutions-as-things-you-sometimes-acquire-on-credit metaphor from a different perspective, which may serve to cast it in a more positive light than usual:
The âpile of technical debtâ is essentially a pile of knowledge â everything we now think is bad about the code represents what weâve learned about how to do software better. The gap between what it is and what it should be is the gap between what we used to know and what we now know.
Thinking of tech debt in this manner feels more like an opportunity gained by learning stuff vs a liability you have to pay off. And who doesnât love a good opportunity?!
You can refuse to take that opportunity if you want, but itâs a tragic waste of your hard-earned knowledge â a waste of the investment you previously made in learning â and eventually youâll be losing money, and losing out to competitors who will be making the most of their knowledge.
Mike McQuaid and Justin Searls join me in the wake of the RubyGems debacle to discuss what happened, what it says about money in open source, what sustainability really means for our community, making a career out of open source (or not), and more. Bleep! đĽ VIDEO HERE đ
Andrej Karpathyâs latest project provides a full ChatGPT-style LLM (including training, inference and a web UI) in ~8k lines of Python. The whole thing is cool and interesting, but Iâm linking to the web UI frontend because itâs a great example of building something super useful with vanilla HTML, CSS, and JavaScript. Sometimes just a few âonclickâ handlers is all you need, ya know?
/via Simon Willison
Boyd Kane:
After 40 years of persistent badgering, the software industry has convinced the public that bugs can have disastrous consequences. This is great! It is good that people understand that software can result in real-world harm. Not only does the general public mostly understand the dangers, but they mostly understand that bugs can be fixed. It might be expensive, it might be difficult, but it can be done.
The problem is that this understanding, when applied to AIs like ChatGPT, is completely wrong. The software that runs AI acts very differently to the software that runs most of your computer or your phone.
Fascinating take by Hillel Wayne, as per usual:
Color carries a huge amount of information. Color draws our attention. Color distinguishes things. And we just use it to distinguish syntax.
Nothing wrong with distinguishing syntax. Itâs the âjustâ that bothers me. Highlighting syntax is not always the most important thing to us.
- Free software hasnât won
- Pipelining in psql (Postgres 18)
- Pwning the entire Nix ecosystem
- FSF announces Librephone project
- On Signalâs post-quantum makeover
- Hyperflask: full-stack Flask + htmx framework
- An autonomous agent for deep financial research
- A text-based work management system for techies
- I am sorry, but everyone is getting syntax highlighting wrong
- Worldâs first decentralized trained open-weight diffusion model
- Environment variables are a legacy mess: letâs dive deep into them
Thatâs the news for now, but we have some great episodes coming up!
- Wednesday: Ellie Huxtable talkinâ Atuin Desktop
- Friday: Kaizen 21 with Gerhard Lazu
Have yourself a great week,
lips of knowledge are a precious jewel,
and Iâll talk to you again real soon. đ
âJerod