How do you test agentic coding skills in interviews? Do you let engineers run a coding agent? Do you ask how they know code+tests are actually correct? Do you see if they can orchestrate multiple agents effectively? If you’re not doing this already—you’re probably doing it wrong.

I’ve been running a multi-agent workflow for design + coding. Instead of one LLM, I set up 4 agents: Fast Iterating Developer, Test-Conscious Developer, Senior Engineer, Architect. It’s not just about code — they generate artifacts like design docs, conflict reviews, and structured PR replies.

"Election Rigging Response Act" Finally some Democrats awake and figure out how to fight back. How encouraging. I am absolutely against gerrymandering - THE biggest flaw of the American election system. But this bill is so tightly tied to the blatant attempt by Texas to rig the midterms.

Had the most frustrating experience with an AI coding agent recently. Every time it fixed a few tests, it would declare "mission accomplished" while dozens of tests were still broken 🤦‍♂️

Spent months fighting my coding agent's tendency to write duplicate, bloated code. Then I realized: stop fighting it, work WITH it. The Two-Phase Approach: Let AI build messy but functional code Use AI to analyze and systematically clean it up

Coding agents are great — until you ask them to shut down a server. My async tool server hit a wall when Claude and AMP failed to handle real-world teardown logic. Read how I debugged it, what the agents missed, and what I learned: 👉 novibe-vibe-coding.hashnode.dev/when-coding-...

Just shipped a post on how I built a custom MCP tool server for Claude Code + Sourcegraph AMP: 🧰 Tool fetches & parses logs 🧠 Agent diagnoses & collaboratively fixes code 🆕 I just review & push Full write-up: novibe-vibe-coding.hashnode.dev/vibe-coding-... GitHub: github.com/mstriebeck/g...

👾 I wired up Claude + Sourcegraph AMP to my GitHub PRs: Pulls review comments Agent generates changes + replies Script posts back to GitHub LLMs as real collaborators, not just autocomplete. Details + screenshots 👇 novibe-vibe-coding.hashnode.dev/integrating-...

I would love it if coding agents could be more intgrated into our (current!) workflow systems, e.g. GitHub: if the agent writes and changes code, I'd love to use the GitHub PR review UI to leave comments like I always do. And then it gets sent back and the agent address it

So, let me get this straight: A week ago Trump gives a speech in Saudi Arabia declaring that the US should stop focusing on “Nation Building” and stop focusing on democracy or not, but focus on how valuable other countries are to the US.

Well done Canada!!! Well done!

Claude Code just told me: “Instead of trying to fix all tests right now, we should let the dev team fix failures later.” That one sentence sums up my biggest frustration with AI coding agents: they treat tests as an obstacle—not as part of the development process.

So, many people in Trumps circle point out how the drop in the stock market is part of a larger plan. In particular to make it easier to get the deficit under control. This talk with Scott Bessent makes a pretty compelling case. But everybody agrees that the path to success here is very narrow.

I wonder if Elon is the new Kryptonite of the MAGA movement. It's incredible to witness how a single person can upset so many people so quickly ... and either doesn't seem to be aware of it ... or worse: doesn't care.

Wow — very disappointed with Claude Code 😕 I use Claude a lot for coding help — it’s usually the best LLM for code-related tasks. So when Anthropic launched Claude Code, I was genuinely excited.

Unintended side effects ... Past US administrations convinced Nato that it would be good if all countries use the same equipment. In particular the F-35 fighter jet. And these are the current active or planned orders: ...

“For more than two centuries, it has been established that impeachment is not an appropriate response to disagreement concerning a judicial decision. The normal appellate review process exists for that purpose,” Chief Justice John Roberts is one of my heroes. He is a shining example of integrity.

I couldn't agree more with Joe Rogan. These days Jon Stewart is one of the very few who makes fun of both sides. Does he have a slant? Yes, absolutely, he's a liberal. But that doesn't stop him from calling out democrats/liberals on their hypocrisy.

This is an issue where I can't see any win-win's. I can totally understand content creators who are rightfully concerned that nobody goes to their content anymore but it's only consumed through the resulting AI products. arstechnica.com/google/2025/...

So, a sitting president advertises openly products from one company over other (American!!!) companies. And nobody seems to care. If this isn’t nepotism then I don’t know what is …

This is an excellent choice. A few months ago, the liberal party was basically dead. But Trump and now this leadership will bring them back. ... it's tempting to thank Trump ...

Just listened to the AllIn podcast. At the end they had a segment with David Sachs who is now working as part of the Trump administration. Apparently, before starting his job, he ...

This is a part of AI innovation that's often overlooked. Focusing on first principles and not just throwing money and GPU's at the problem. And Anna Patterson is one of the few people who can pull this off! www.bizjournals.com/sanfrancisco...

Decision like these are super tough. But I’m proud to work at a company that always puts users and user experience first. We’ll get there … it will just take a little bit longer ... and it will be awesome!

Tesla sales year-over-year: - down 46% in Europe (76% in Germany!) - down 49% in China (while competitor BYD increased by 90%!!!) - down 73% in Australia Being America-First (and with that all other countries second) and supporting right-winged opinions has a cost.

A few years ago, there were two opinions about DEI: - one side said that it’s important to focus on and makes the business better - the other side said that it’s just pandering

Very funny: when I created my last post (bsky.app/profile/mstr...), I ran it by Claude for any feedback. Claude made some changes and listed and explained all changes explicitly. All changes except changing “(GitHub Copilot, Cody, Cline…)” to “(GitHub Copilot, Cody, Claude…)” - that wasn’t listed.

Fascinating to consider where AI coding tools (GitHub Copilot, Cody, Cline...) are taking us. From my experience (and recent discussions with others), they're good but require careful oversight—both in crafting prompts that generate high-quality code and in reviewing/refactoring the output.

Over the past couple of months, I have used ChatGPT, Claud and Gemini as well as some assistants on top of them (Cline, Autopilot) to write code. At this point, they seem to be the equivalent of an incredibly smart and knowledgeable but very junior engineer:

I very much agree with this article. Upon scrutiny, many of the The Free Press articles don't hold up. I love the idea of The Free Press - but right now, they are often as biased as the media that they want to replace.

When I joined Apple, I needed to learn Swift. I decided to use a personal project (an LLM-based news reader) that I worked on for some time and use LLM’s (ChatGPT, Claude and Gemini) to help.

Elon Musk launching DOGE accounts for all agencies compresses all my opinions about this effort: The Good, The Bad and The Ugly:

If you are working in or are interested in the current state of AI, then you don’t want to miss this episode of Lex Fridman’s podcast: www.youtube.com/watch?v=_1f-.... The two guests Dylan Patel (@dylan522p) and Nathan Lambert (@natolambert.bsky.social) have broad knowledge and are in the trenches.