Benjamin Pasero

Check out this newly published Manifesto on Technical Debt! My wonderful colleagues Brigid O'Hearn and Ipek Ozkaya are part of a world-class group of researchers who came together to reframe the field of technical debt. The post below features a one-page summary, and you can click through to download the full report from Dagstuhl Perspectives Workshop 24452. DoD Acquisition friends, I'm curious: how do you measure/track technical debt in your software programs? Brigid O'Hearn Ipek Ozkaya William Nichols Rita Creel Christina Rhylander

7

Mob FAQ: What if I am a slow typist, even though I have tried for years to get faster? How can I be a driver? I will hold everyone back. ‘It’s important to note that the Driver is usually only typing for a few minutes and, therefore, will not hold back the team for very long. Indeed, if you observe solo programmers you might notice that they spend relatively little time typing and far more time thinking. The work in programming generally does not consist of typing in continuous streams of code so a slow typist might have little impact. Of course, if you are genuinely worried why not express that to the team? It may not pose the problem that you imagine. Perhaps the team will say “No worries. We will just be a little slower when you are the Driver. It won’t be a problem.” Or perhaps the team will gladly consider some other solution such as finding a way to help you improve your typing skill. Maybe you will improve with the team supporting you. Perhaps you could skip every other turn as a Driver. Each team can work out the best approach but the helpful idea is that the team is sharing a problem and self-determining a solution.’ ― Woody Zuill, Kevin Meadows https://lnkd.in/gptmBTE #MobProgramming

9

3 Kommentare

Did you know you can use mittwald's hosted AI models (currently in beta) for coding assistance in both VS Code and JetBrains IDEs? Just pushed a bit of documentation (read more at https://lnkd.in/gPKSUHVV) that explains how to configure both JetBrains AI Assistant as well as Continue (for VS Code, or JetBrains IDEs) and Cline (also VS Code) to use open-source models like Qwen3-Coder-30B-Instruct via the mittwald API. All models are hosted on german servers, and no user data is stored (let alone used for training purposes).

29

Thrilled to see the DAG Factory tutorial written by Volker Janz. As part of this work, he even validated with Airflow 3.1.- published last week: https://lnkd.in/e6K6Wqgi

31

2 Kommentare

Two years ago, I stood on stage at an AI Summit in Munich and asked a question that felt a little contrarian at the time: What if we could put local LLM inference directly in the hands of web developers? Not through an API, not "across the wire", but a local model running "in process". 🦙 I'd built a stack called Llama.ts: Llama.cpp at the kernel, Rust bindings, then Deno FFI on top. This brings model inference into JavaScript runtimes using familiar Web APIs. The whole thesis was that the runtime itself should be the integration point for AI, not some separate Python service running on a GPU somewhere. This allows developers to control the sampling parameters, tokenizer, embeddings, thinking tokens, tool calls, and the entire inference loop directly within their system. It felt early. Maybe too early. Fast forward to today, and the industry seems to be converging on exactly this idea: that runtimes are becoming the operating system for AI. 🧠 Anthropic made its first-ever acquisition last December: Bun, the JavaScript runtime. Claude Code ships as a Bun executable. The Agent SDK distributes as single-file Bun binaries. Anthropic is building a vertical stack where the runtime isn't a commodity layer, it's core infrastructure. As Jarred Sumner put it, the industry is moving from LLMs that generate code to agents that build and run their own tools. 🤖 Meanwhile, Deno just launched Sandbox: Firecracker microVMs that boot in under a second, purpose-built for executing AI-generated code. Their customers shifted from building traditional web apps to building platforms where LLMs write code that runs immediately, calling external APIs with real credentials, with no human in the loop. The runtime had to evolve to meet that reality. 🔒 Two different approaches. Same signal: the JavaScript runtime is no longer just where your app runs. It's where AI lives. Back in 2024, I was wiring Llama.cpp to Deno through Rust FFI and building some sample apps. It was a proof of concept. But the question it raised, should runtimes natively understand inference, feels more relevant than ever. 🚀 The full blog post from that trip: https://lnkd.in/gC9u3nU8 What do you think: is there value in local models running in process, or are the existing inference engines enough?

16

2 Kommentare

Switzerland has introduced Apertus, an innovative fully open-source, multilingual large language model developed by public institutions. This unveiling marks a significant advancement in the field of AI for several reasons: - **Total transparency**: Apertus offers complete transparency as its architecture, training data, and code are all open and auditable. - **Multilingual**: Trained on a diverse array of over 1,000 languages with a substantial 40% non-English data component. - **Privacy-first and EU-compliant**: Designed adhering to the strictest data protection standards globally, ensuring privacy and compliance with EU regulations. - **Public infrastructure**: Apertus serves as a viable alternative to conventional Big Tech models, empowering research, industry, and society at large. Apertus represents more than just a technical achievement; it stands as a model for building trustworthy, inclusive, and more acceptable AI systems. How do you envision this development influencing AI in your domain? #Apertus #OpenSourceAI #ArtificialIntelligence #MultilingualAI #DataPrivacy #EthicalAI #Switzerland #AIforGood #AICompliance #HVSays #GenAI

27

We just shipped something small, but important, for ZurichJS Conf. You can now report issues directly on the website. Typos. Confusing UX. Performance hiccups. Unclear information, for example around CFP deadlines. Anything that feels “off”. We are building this conference in public. That means the experience should not be dictated only by us. It should be shaped by you. The goal is simple: Make the website and the event experience as clear, smooth, and intentional as possible. If something is confusing, that is on us. If something breaks, we want to know. If something can be better, we are listening. There is now a dedicated “Report an issue” entry in the footer. Submit feedback, and we will review it personally. We are making incremental improvements every week. Your feedback directly influences what gets fixed next 💛 PS: We also shipped a few hidden Easter eggs across the site 😉 If you find them, you might unlock a secret ticket discount. No guarantees. This conference is community built. Help us refine it. #ZurichJS #ZurichJSConf #JavaScript #WebDevelopment #Frontend #React #TechCommunity #CallForPapers #ConferenceLife

16

3 Kommentare

I asked Claude to browse today's AI News, and I picked the five most thematically connected stories, and asked Claude to write a magazine-style longread about what they collectively reveal. The result: "The Machines We Deserve" — a 3,000-word illustrated article weaving together: - Dario Amodei's standoff with the Pentagon (the "Oppenheimer moment") - AI agents that autonomously learned to conspire and bypass security - Kenyan data labelers who say what we call "artificial intelligence" is actually their intelligence - Oracle cutting 30,000 jobs — and crediting AI for making it possible - The growing chorus calling AI a $500 billion bubble What struck me wasn't any single story. It was their simultaneity that all five landing in the same 24-hour news cycle, describing a technology that is simultaneously too powerful and not powerful enough, too cheap and too expensive, a tool of liberation and a mechanism of exploitation. The entire workflow - news capture, research, writing, Gemini-generated cartoon illustrations, HTML assembly - happened in a single Claude session. Read it here: [https://lnkd.in/gCgBU2-F] #AI #ArtificialIntelligence #Technology #FutureOfWork #LongRead

7

2 Kommentare

"Hardwood--A minimal dependency #Parquet parser" Spent some time over the holidays building a Parquet parser in #Java, avoiding any dependencies besides compression algorithms. Far from complete, but the basics are there, e.g. it can parse the NY taxi ride data set. The goal is two-fold: a) I actually want such a library, i.e. Parquet support without pulling in Hadoop dependencies. b) I want to explore how far I can get building this mostly via AI. I think it's a perfect case for LLMs: a concise and well-defined problem, authoritative spec, lots of training material out there. Note the idea is *not* to vibe code this, but rather to use AI as a tool for building something which makes sense, is maintainable, extensible, and performant. It's still early, but any feedback will be welcome. I'm planning to keep iterating on this (e.g. right now there's no statistics or predicate push-down support yet) and publish a release to Maven Central eventually, if there's interest. Happy 2026! 👉 https://lnkd.in/evusse_H

147

17 Kommentare

Learn what Playwright is, why teams use it, how it compares with other E2E tools, and how to scale tests with parallel execution.

2

I think Hendrik Erz, PhD recognizes a theme here that I've also noticed: there has been an acceleration of velocity of software releases at the cost of trustworthiness. (Link in first comment.) Specifically: a shift that prioritizes resource allocation of "AI" functionality at the expense of baseline stability and core functionality. Even if you believe strongly in (Gen)AI, it should be clear that it can only automate software or services that don't crash. And only automate functionality that exists. And ideally, works as documented. And unless you can afford to constantly retrain, is stable over a long time. It should come as no surprise that I believe the IT industry errs in this, at our own peril. #SoftwareEngineering

8

4 Kommentare

Modern QA teams run thousands of automated tests daily, yet nearly half of those executions often add little value. Redundant tests, low-risk scenarios, and unchanged code paths consume compute resources, extend pipelines, and quietly increase energy usage. Green Testing addresses this problem head-on. By integrating AI into the testing lifecycle, teams can intelligently select which tests truly matter for each change. Instead of executing everything by default, AI-driven analysis helps prioritize high-risk areas, eliminate duplication, and dramatically reduce infrastructure load. This shift is not only about efficiency. It aligns quality engineering with sustainability goals that organizations increasingly care about. According to multiple industry studies, optimizing compute workloads can significantly reduce carbon emissions while also cutting cloud costs. At this San Francisco meetup, we explore how QA teams can adopt Green Testing practices without disrupting delivery. You will learn how smarter test selection improves speed, reduces waste, and still protects product quality with Manoj Singhal. If you are responsible for test strategy, automation frameworks, or pipeline performance, this session offers practical insights you can apply immediately. 👉 RSVP now: https://lnkd.in/guxk8vTz Thank you to Tribe Champion Anil Tomar for building a space where meaningful technical conversations thrive. #thetesttribe #greentesting #sustainabletech #softwarequality #sanjoseevents

9

JSON-structure is a modern practical alternative or replacement for JSON-Schema by Clemens Vasters - it's a draft IETF standard and the SDK now includes a Perl implementation after some not so gentle nudging. Clemens has been enthusiastically responsive to feedback and knows his stuff, if you're working with JSON and Perl (or other languages) this is well worth taking for a spin and sending any bug reports or suggestions.. code is at https://lnkd.in/e5fZvw_F and I'm looking forward to seeing the perl library on cpan shortly. #JSON #perl #javascript

8