Finally. Thirteen years after the RX1 was first released, and ten years after the RX1R II, Sony has released a video announcing an update to one of my favorite cameras. Packing the 61MP sensor, focus, and image processing from the A7R4 into a ultra compact body with the same lovely 35mm/f2 lens of its predecessors, it’s pretty much pitch perfect.

Now, some people are complaining about the price. At $5098 in the US, it’s pretty steep. But the RX1 was never cheap. The RX1R II was priced at $3298 ten years ago, which felt really steep. That's $4333 in 2025 dollars with inflation. “OK,” you say, ”that’s cheaper, right!” Oh no, my friend. You’re forgetting about the new US import tarrifs of 25% on the declared import value. You’d have to make several assumptions to calculate a true adjusted cost, but... it’s close.

So, no, the camera isn’t cheap. The value proposition won’t make sense for many. But, this was never meant to be a mass market camera and the price is a rational extrapolation factoring in inflation and tariffs. If it made sense to you 10 years ago, it makes sense now. If it didn’t then, it still doesn’t. 🤷

Anyway, I'm stoked about this new release of an old favorite.

Rands (aka, Michael Lopp) has been on his own deep dive into the world of the robots (aka, AI) and how they affect the future. There’s good news and bad news, he writes, and we’re in that inflection point where we can’t yet grasp the long-term implications of the tools that are suddenly at our disposal.

Like social media before it, we can’t conceive of the long-term implications of these tools suddenly being at our disposal. We’re either giddy with excitement about their potential or digging deep to find any and all reasons not to trust them. But trust your instincts.

Maybe the most reasonable thing right now is to be both giddy and skeptical as hell. Go all in, and yet trust very little. We thought we knew what the Internet would bring us in the mid-to-late 90s, and we weren’t even close to knowing it all. We got some things right about five or ten years too early, and yet made ludicrous guesstamates about what might be possible that now seem juvenile.

Brie Wolfson travelled to Kevin Kelly’s studio in Pacifica to have a chat about how to have impact in the world and be happy about it. Their conversation explores alternatives to the Silicon Valley mantras of maniacal focus and relentless pursuit, advocating instead for a career full of creativity rather than a linear climb up a ladder. In case you don’t know Kelly, he helped start WELL, one of the first online communities, in the mid-80s and was co-founder of WIRED in the 90s.

I asked Kelly about the tradeoffs of focusing on a single thing if you want to be great... “Greatness is overrated,” he said, and I perked up. “It’s a form of extremism, and it comes with extreme vices that I have no interest in. Steve Jobs was a jerk. Bob Dylan is a jerk.”

As someone who has certainly had an interesting path through life, bouncing around between software, writing books, and photography, I love this permission to work differently. To measure success by internal satisfaction rather than just net worth or employee count. There's value in being a generalist who follows their interests, even if it means abandoning things when they no longer serve you.

If you were an avid reader of WIRED in the 90s — you know who you are — you’ve read Peter Leyden’s work. He’s now working on a series of talks and essays called “The Great Progression: 2025-2050.” His thesis is that the arrival of AI, the rise of clean energy technologies, and the emergence of biotech add up to one of the great transitions of history. A transition that’s on par with what we saw in the post-war era, the gilded age, or even the founding of the United States. And, interestingly, each of these transitions also came with the kinds of social upheaval we see today.

Building on an existing partnership, Anthropic worked with Rick Dakan and Joseph Feller to create a free 12-lesson course around Dakan’s and Feller’s AI Fluency Framework. The framework consists of four interconnected competencies: delegation, description, discernment, and diligence, and emphasizes human-AI collaboration over just understanding AI technology.

Even if you’ve been working with AI for a while, I think it’s worth spending the 3-4 hours the course takes to hone your thinking about how to work effectively with AI.

Martin Fowler thinks that working with LLMs represents as fundamental a change in software development as the shift from assembly code to higher-level languages like Ruby and Fortran was. But unlike those transitions, this isn't just a jump up in abstraction level. Martin writes:

When I wrote a Fortran function, I could compile it a hundred times, and the result still manifested the exact same bugs. Large Language Models introduce a non-deterministic abstraction, so I can't just store my prompts in git and know that I'll get the same behavior each time. As my colleague Birgitta put it, we're not just moving up the abstraction levels, we're moving sideways into non-determinism at the same time.

I think that how much of an impact this has will depend on where you use LLMs in your software development process. When you're working with an LLM to create code, but then committing the results to be used, the line of non-determinism stops well within your view. On the other hand, if you're using prompts to write code and not looking at the output, then you'll have to sort out what systems you need to provide accountability that the final result meets your requirements, even if you're not able to inspect every line of code.

I was a bit skeptical when Warp started adding AI features to their terminal app. I’ve really enjoyed watching Zach Lloyd and team work on their take on a modern terminal over the last few years, but using AI to turn "find files with X" into grep commands felt pretty underwhelming. Luckily, that was just a stepping stone and Warp has now evolved into what they’re calling an agentic development environment.

Now with the release of Warp 2.0, it’s clear that they’re chasing the same general idea that Claude Code is, but they’re building the functionality into the UI of the terminal itself:

Because Warp is the platform and not a CLI-app running within the terminal, Warp has much richer UX capabilities than any CLI agent possibly can. For instance, unlike in CLI coding agents you can edit diffs directly in Warp’s native code editor; no context switching to an IDE required.

Furthermore, they’re taking advantage of the fact that they can use different models from different providers to do things like use OpenAI‘s o3 model for planning, and Claude 4 Sonnet for code generation. It’s an interesting take, and one I’m looking forward to checking out and seeing how it fits into this broader trend of rethinking our development environments.

In his keynote at YC’s AI Startup School earlier this week, Andrej Karpathy talks about how software development is changing with what he calls a third paradigm of programming. The first he calls Software 1.0 and is composed of normal code. Software 2.0 is the weights in a neural net that categorize things in a fixed-function kind of way. Software 3.0 is where the neural nets become programmable with English. In order to succeed, we’re going to have to be good at using all three of these paradigms together.

Philip Carter has started calling LLMs “weird” or “inverted” computers because of how differently they behaive compared to the things we’ve been calling computers for almost a hundred years.

We've spent decades working with computers that are incredible at precision tasks but need to be painstakingly programmed for anything remotely fuzzy. Now we have computers that are adept at fuzzy tasks but need special handling for precision work. I genuinely do not know what to make of this yet!

They are different beasts, to be sure. Which is why I’ve been spending so much time lately really getting my head around how to use them.

Austin Parker, Honeycomb’s Open Source Director, started with a standard Honeycomb demo showing some interesting latency spikes and then ran it through an AI agent he wrote in a few days integrated with their new MCP (Model Context Protocol) server. The agent took 80 seconds, made eight tool calls, and sorted out what was going on. All for the price of $0.60 of tokens.

This is a seismic shift in how we should conceptualize observability tooling. If your product’s value proposition is nice graphs and easy instrumentation, you are le cooked. An LLM commoditizes the analysis piece, OpenTelemetry commoditizes the instrumentation piece. The moats are emptying.

This isn’t something he despairs about, however. He thinks these productivity increases will bring more software which will need more tools. Of course, we have no idea what that’s going to look like yet, but I love this positive spin. Nobody really knew what the web was going to do to the world thirty years ago, just like nobody knew what the automobile would do to the world over a hundred years ago. The question isn’t whether this transformation will happen, but what we'll build with these new capabilities.

Researchers at Apple investigated whether Large Reasoning Models (LRMs) like o1, o3, DeepSeek-R1, and Claude 3.7 Sonnet truly reason or simply pattern-match using longer chains of thought. They found that for medium complexity tasks, thinking models show advantages over their regular LLM counterparts, but for high complexity tasks, both model types show significant limitations. This suggests the “thinking” capability may be more limited than they appear.

Of course, this doesn’t change the usefulness of these models, but better understanding how they work — and which models are good for which tasks — is essential for using them well.

Cloudflare open sourced their OAuth provider library for Cloudflare Workers, which was largely written with Claude. Kenton Varda writes in the README.md for the project:

I was an AI skeptic. I thought LLMs were glorified Markov chain generators that didn't actually understand code and couldn't produce anything novel. I started this project on a lark, fully expecting the AI to produce terrible code for me to laugh at. And then, uh... the code actually looked pretty good. Not perfect, but I just told the AI to fix things, and it did. I was shocked.

Kenton documented the process by including the prompts he used and these show a partnership between the AI and developer. Almost every feature required iteration and refinement, but in much the same way as a senior developer talking to a junior. Kenton says as much in a post on Twitter, er, I mean X:

A lot of people seem to be looking at my Claude Code prompts to try to learn some sort of techniques. Folks, just to be clear, I didn't know what I was doing, I just talked to it like I'd talk to a human.

This matches my own experience. Pairing with LLMs is the whole ballgame right now.

Claude Code began as an internal tool inside Anthropic originally created by Boris Cherny. In this conversation, Claude Relations lead Alex Albert talks with Boris about how Claude Code started, its incredible adoption curve inside Anthropic, what Claude Code has unlocked, and some tips and tricks for using it.

I've been a fan of Claude Code since the moment I first saw it and am really happy that Anthropic pursued its public release.

For the last year or so, I’ve been watching the development of Zed with a lot interest. Instead of being built on a web browser like Visual Studio Code, it’s built on Rust and is insanely fast. The downside is, however, is that Cursor or Windsurf, both of which are forks of VSCode, have been the go-to place to work with LLMs in your code.

No more. The Zed team released their new agentic coding feature today, and it’s really good. Is it better than the agent experiences in Cursor or Windsurf? I wouldn’t go that far. But Zed is in the game now. Combined with the rest of the Zed experience, it means we have a choice that’s not implemented on a web browser and not subject to having core functionality disabled by Microsoft over licensing issues because it’s a fork of VSCode.

Earlier this month on the Tailscale blog, Avery Pennarun started out the year by looking at how far we’ve come in computing over the last twenty years. Compute is 200,000x faster. Web servers are at least 100x. We can put 12TB of RAM into a server. Hard drives are 100x bigger and last longer. And, the c10k problem is now the c10000k problem.

It’s a thought-provoking piece showing how we can do more than ever, with less than ever. Are we keeping up with our tools?

Reading Avery’s post injected some optimism into my day. I needed that.

In the early 2000s, Cameron Moll was working on a presentation about problem solving and wanted to visualize Abraham Wald‘s work in World War II studying the bullet holes in aircraft returning from action. So, he took a diagram of an airplane and drew red dots on it to represent the damage that was visible on surviving aircraft. That became the de-facto illustration for survivorship bias.

I think the most fascinating part of this story is that Wald’s original report from 1943 was un-illustrated. I had always assumed that the diagram was from the original work, and not something that came 60 years later.