Dagens Vibes — 21. juni 2026

Deedy@deedydas

Most software engineers are facing an identity crisis bordering on depression. As CTOs aggressively evangelize tokenmaxxing, a class divide ensues. The lazy. The lazy push code. They don't write it. They don't manually test it. They don't even read it. They're on autopilot. See Jira ticket, prompt for task, submit code. Many of them are barely on their computer the whole day. A comment on the PR asking why they did this? The lazy ask AI. A Slack message? The lazy ask AI. Need to prepare for standup? The lazy ask AI. As long as it sounds enough like them and isn't detected. Some of the lazy are even overemployed, and work multiple jobs. The lazy smart ones get away with this, and even rewarded. After all, software engineering for the lazy is just a dance to convince your colleagues you're smart and hard working. The craftsmen. The craftsmen are tired. Very tired. 15 PRs in queue. Slack blowing up. The entire burden of review falls on the craftsman. The burden of understanding. They try. They work their way through the code, thoughtfully commenting to improve what ships. The response? A lazy: "That's a clever idea! You're absolutely right." with an incorrect change. It's fine, the craftsman says. I can fix them. They write a doc urging his colleagues to be better. The next day? 20,000 line PR to review. Day after day, their workload grows. Bugs seep into production. No one seems to care. Another round of AI is thrown at it. Their animosity to their colleagues rises. Eventually, they give up. It's just not what it used to be. The craft they loved is dead. They eventually wake up, a lazy. This isn't all companies. Many companies are genuinely more productive, adopt the right set of principles and practices around AI development and have highly talented teams that trust each other. It tends to happen in bigger companies that are 10+yrs old with a higher talent variance. But it happens. A lot.

Sudo su@sudoingX

the DGX Spark nvidia sent me is a full supercomputer that fits on my desk. GB10 grace blackwell, 128GB unified memory, sips power off a wall socket, runs models that needed a server rack two years ago. spent last night pushing it on the hardest case i could find, a dense 27B model, to map exactly where it flies and where physics bites. all measured. > 1. dense decode is memory bound everywhere, not a spark thing. to write one token the gpu reads all 27B weights out of memory, and the spark's compute is so far ahead of its bandwidth that the chip sits half idle waiting. baseline 7.64 tok/s. hold that number. > 2.this is where the spark shines: speculative decoding. the model guesses a few tokens ahead, the spark verifies them in one batched pass, one weight read confirms ~4 tokens. 7.64 to 17 tok/s, a clean 2.2x, output byte-identical, one flag. it works precisely because the spark has spare compute sitting idle, and this finally puts it to work. the headroom you paid for earns its keep. > 3. honest catch: that's a short-context win. at 256k tokens it fades to 1.37x because you cross from memory bound to compute bound on attention. physics on any box, not the spark. > 4. the payoff, the spark flexing: switch to a same size MoE and it does 21.7 tok/s at that same 256k, nothing special turned on. a 256k context window at usable speed on a box next to my coffee. that's what nvidia is putting on developers' desks. the lesson: match the model to the machine and the spark is a beast. dense plus spec decode for short sharp work, MoE when you need the long context. either way it's a datacenter's job running off a wall socket. bookmark it, all measured on one box. (nvidia sent me the spark, no money changed hands, every number is mine)

GREG ISENBERG@gregisenberg

GLM 5.2 might be the “ChatGPT moment” for local AI The moment many of us see the value in local models GLM 5.2 isn’t perfect, but it’s really good 1M token context window so it holds an entire codebase at once. Top open model on coding right now. MIT licensed with zero restrictions. And it runs on your own machine through Ollama or LM Studio. When you're using it, it kinda feels like Opus 4.8. I honestly can't believe how good it is. 2027 is probably the year of local AI.

Thorsten Ball@thorstenball

Under-discussed: how good agents are at debugging issues in production. Everybody's constantly talking about how well they write code, but give an agent access to the gcloud CLI and a screenshot of a graph and, good god, will it go.

Dan McAteer@daniel_mac8

This is one of the coolest open-source AI agent projects I've seen in a while: 'Understand Anything' It's a plugin for Claude Code, Codex, OpenCode etc. that analyzes your codebase and turns it into a knowledge base that you can interact with. It explains the codebase to you, rather than showing you the structure. It seems like it's designed for code but I opened my Obsidian vault of podcast highlights in Claude Code, then ran /understand. The result is a knowledge graph that I can search of highlights from 888 podcast episodes and 144K lines of markdown text.

Gergely Orosz@GergelyOrosz

One interesting trend: I’m seeing *so many* VC-funded, internally built infrastructure and bootstrapped solutions around “building a context layer for engineering teams.” Aka trying to solve the problem of “if only eg Claude Code had the context from all your other systems”

Peter Yang@petergyang

You can now make videos like this using HTML. I didn’t believe it at first until the HyperFrames team showed me how they do it directly in Codex and Claude Code. Best of all, the tool is 100% free and open source. But you need to learn the right workflows to make the videos look great - including generating a frame.md, storyboards, and more. 📌 Subscribe to get our full tutorial tomorrow: https://t.co/NSTj7ijxE0

dex@dexhorthy

there is exactly one word you should care about when designing any AI coding system its not loops its not harness engineering its not spec driven dev its not software factory the word is LEVERAGE - if you’re not designing your process so you can make decisions EARLIER, more cheaply, then you’re not giving yourself more leverage.

Howaboua@Howaboua

Another @herdrdev update. TLDR you should really try it and most definitely install the skill. I combined herdr skill with my "try-rs" skill for temp dirs (https://t.co/ebE3ddz272) and it's now testing my Claw extension onboarding flow e2e inside of durable test folders. What feels SUPER good is that using tmux, Clankers had some serious issues prompting Pi TUI (they would put in the prompt, not understand how to insta-send it, weird line breaks etc), and trying to guide them was always a downward spiral into making it worse. Herdr does it first time, correctly. Admittedly I never had a tmux skill, assuming it's been baked well enough into weights. Still, night and day difference.