François Bouteruche

Offering a developer portal is essential in the #FastFlow journey. In my latest article, I discuss why adopting a product mindset is crucial for creating a seamless interface between stream-aligned teams and enabling/platform teams. #TeamTopologies #Accelerate #PlatformEngineering

16

Why context matters when you're a developer (and why #DDDesign is a very practical approach) Thanks to Guillaume C. for the link 😄 https://lnkd.in/eZbdz-BT

7

"AI in TDD" Thank you Valentina (Cupać) Jemuović (Cupac) and Laurent P. for this "AI in TDD" top knowledge ! KEY TAKEAWAY: "The design produced by AI was in my mind sometimes ... sometimes OK but AI needed human prompts, AI needed a human who understood good design principles so that when AI did not understand it well - to be able to prompt it. Now, when prompted it with more guidance (by human) - YES, AI was able to improve the design. " - Valentina (Cupac) Jemuovic " AI in TDD Agenda: 1. TDD - Overview of the TDD Cycle (s) 2. AI in TDD - Using AI for the TDD cycle 3. COPILOT DEMO - TDD with Copilot (Java) 1. TDD - Overview of the TDD Cycle (short) RED-GREEN-REFACTOR "... YES, AI was able to reach THE GREEN STATE after iterations (something that avarage developer could also do ... AI is here the most helpful)" "... sometimes when I asked "can you refactor the code" - AI would then ask me to be more specific, AI would ask me "WHAT DO YOU WANT TO ACHIEVE WITH THE REFACTORING ?" ... and then I had to maybe choose some options ... "with refactoring it can be ... " and AI may suggest ideas and those may be good OR may not be good. BUT, if you guide AI - AI CAN EXECUTE REFACTORING ... but it is just the idea of WHICH refactoring to do." 2. AI in TDD - Using AI for the TDD cycle Discussion (strenghts and weaknesses) Key takeaway: using AI assistant - developers need to be really thoughtful and critical 3. COPILOT DEMO - TDD with Copilot (Java) Practical coding demo using COPILOT (user story, acceptance criteria, using copilot to implement user story with TDD) github com / valetinacupac / tdd - copilot - java " Ref: "AI in TDD by Valentina Jemuović" https://lnkd.in/d5gVfBhu

18

Web apps can be simple I created a "proof of concept" application for a friend. And I am amazed at the fact that my understanding of "low complexity application is the norm" actually is true. The problem domain is fairly complex, and the security requirements very stringent. But I have been able to put together a foundation that will allow her to get help from just about anyone in the world to keep her application updated. In my opinion, it cannot get more #lowcode than this. Anything else, is just lock-in. And will come back to bite people very quickly. By the way, I had a similar experience when I built the same app, using Java, H2DB and the internal HttpServer that ships with Java. The only external thing I needed to use was the Thymeleaf templating engine, and I think it was an overkill. Should have just relied on Java Server Pages. Yes, you heard that. No Kubernetes. No microservices. No microfrontends. Just plain simple HTML (server rendered). Scales for most users, and keeps the costs low. More details in the blog.

8

1 Comment

You are trying to setup a Symfony setup with AWS Lambda? Let me tell you a little bit about trusted proxies and X-Forwarded-Prefix, just in case you want to run it in a subpath.

2

1 Comment

"Which domain should own this domain event? It doesn't feel like our team should own it..." An extract from one of the many interesting conversations I was part of today at PayFit France. I liked this one because it's a good example of socio-technical architecture. In order to shape the best software boundaries and interfaces, we have to think about how it will impact the teams building the system: 1. which team has the best knowledge of this concept? 2. which team works on features most related to this? 3. which team's cognitive load will be impacted the most by this? 4. which team will be incentivized to build the best implementation? 5. which team works closest with the relevant domain experts? This isn't the only important aspect, but it is an important one. #socioTechnicalArchitecture #softwareArchitecture #ddd #domainDrivenDesign #eventDrivenArchitecture

87

4 Comments

Sylius 1.13 is here and contrary to popular belief, this is quite a lucky number and a great patch for our solution 😀 What's new in this release? 📌 Our API is now fully stable for both admin and shop, with many additional API endpoints. 📌 We are more closely aligned with the Symfony and PHP ecosystem; for example, we've switched from Winzou State Machine (which served us for years) to Symfony Workflow. 📌 The Price History (Omnibus Directive) is now part of our core. 📌 We've introduced several improvements to the Developer Experience (DX), including PHP8 attributes and interfaces for smoother service declarations. 📌 Numerous performance enhancements. You can check out the full changelog here: https://lnkd.in/dAydy9bp A huge thanks to our team and all contributors! 🖖 PS. On top of this, we've also released the 15th patch for version 1.12 😊

21

🎉 I'm super excited to be presenting at apidays in Paris on December 3-4-5 on the brand new JSON Schema conference track: https://lnkd.in/dMtc6ing. 📚 I'll be talking about my learnings working with organisations that successfully maintain large repositories of schemas (>1000), and the common patterns I discovered. The idea is to discuss the benefits of applying software engineering principles to schemas and treat them as you would treat the rest of your code. 🤓 Expect to learn lots of practical tips, and demos of tools you can use to enhance your experience working with schemas, including my brand new JSON Schema CLI (https://lnkd.in/dBXKVm_r). All of it very applicable when working with JSON Schema in the context of API specifications like OpenAPI Initiative, and AsyncAPI Initiative. Hope to see you there! If you are going too, reach out, and huge kudos to Edward Rossa for making this happen!

23

1 Comment

Why I like #VueJs ecosystem 🥰 Time to share some useful tools we use at ERA 2140 💻 >>> #developers : What if you could use any icon from any icon set in you project using a single line of code? Check this: https://iconify.design Installation and usage of iconify in Vue.js is very easy using this wonderful library (unplugin): https://lnkd.in/efukNup9 Note that iconify is compatible with or without any framework, React, Svelte or even pure SVG/css Have fun!

15

Low code / No code enables business users to create their own solutions. They're called "𝗰𝗶𝘁𝗶𝘇𝗲𝗻 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀". They play a 𝗰𝗿𝘂𝗰𝗶𝗮𝗹 𝗿𝗼𝗹𝗲 in companies. Here's how : 1️⃣ Acceleration of application development 2️⃣ Quick response to their own needs 3️⃣ Reduction of shadow IT 4️⃣ Integration with existing systems

7

Claude 3.5 can be used for some incredible use cases. As a developer, the new Artifacts feature is game changing for React. Have a look at these examples.

7

#Monday morning #Copilot prompting triumph 💪🏼😊🍵: In Copilot App for Teams, I pasted a prompt shared by a peer, "Based on my emails and Teams chats from the last 7 days, organize my tasks along the Eisenhower matrix for the coming 15 days." Then, first turn back to Copilot, "Organize this Eisenhower Matrix list into a table." And finally, "Which of these tasks are you able to assist me with and how?" Now, I have my prioritized, categorized list and suggestions for how Copilot can help me tackle them one-by-one. #BringItOn #ReadyForTheWeek #MicrosoftCopilot

28

3 Comments

How can 𝐓𝐞𝐚𝐦𝐬 𝐑𝐨𝐨𝐦𝐬 help 𝐂𝐨𝐏𝐢𝐥𝐨𝐭 🤖 and 𝐀𝐈 𝐑𝐞𝐜𝐚𝐩🧠? Now in 𝐏𝐮𝐛𝐥𝐢𝐜 𝐏𝐫𝐞𝐯𝐢𝐞𝐰 for ALL Rooms! Microsoft Teams Rooms support 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐭 𝐒𝐩𝐞𝐚𝐤𝐞𝐫, to identify the who said what in a room in a transcript to help 𝐂𝐨𝐏𝐢𝐥𝐨𝐭 𝐚𝐧𝐝 𝐀𝐈 𝐑𝐞𝐜𝐚𝐩, for great summaries and prompts. While not every Teams Room has the 𝐈𝐧𝐭𝐞𝐥𝐥𝐢𝐠𝐞𝐧𝐭 𝐒𝐩𝐞𝐚𝐤𝐞𝐫 certified hardware yet, We are 𝐞𝐱𝐭𝐞𝐧𝐝𝐢𝐧𝐠 𝐭𝐡𝐢𝐬 𝐜𝐚𝐩𝐚𝐛𝐢𝐥𝐢𝐭𝐲 𝐭𝐨 𝐞𝐱𝐢𝐬𝐭𝐢𝐧𝐠 𝐡𝐚𝐫𝐝𝐰𝐚𝐫𝐞. It's important to note that the quality may not match that of an intelligent speaker certified device. Therefore, it's crucial to evaluate the advantages of incorporating an intelligent speaker, especially in crucial spaces where attaining the highest quality transcription and attribution is vital. Learn more in our NEW documentation on how to get the best out of Teams Rooms and Copilot today: https://lnkd.in/eYMjxY7F Parisa Taheri, Ph.D. Ilya Bukshteyn Sandhya Rao Kelly Johnson, MBA Alexander Ivanov Jimmy Vaughan Michel Bouman Kyle Spiess Vishnu N. Stav Yagev Igor Abramovski

101

3 Comments

Open Source is the most effective collaboration model we have where competitors and unrelated needs come together to build the future.

1

1 Comment

Restructuring a French software company must be an absolute nightmare: while it’s not hard to see how it all went wrong (having been there, seen that, I estimate the average developer’s productivity to be between 20-30% that of a Silicon Valley one, best case scenario - accompanied by that adorable mix of Gallic arrogance and insouciance), it’s almost impossible to see how one could right the sinking boat: French labor laws make it all but impossible to fire even the most incompetent and lazy of workers, and large scale layoffs are a complete no-go. The real mystery is why would anyone invest there (and, in fact, investors will be “all but wiped out under the current restructuring plan”); the real question being: has anyone learned their lesson here? Atos Picks Onepoint Over Kretinsky for French Firm’s Bailout  https://lnkd.in/gSZ_Qgq2

2

2 Comments

How to build your own Large Language Model ? There are 3 techniques to adopt : prompt engineering, retrieval augmented generation (RAG), and fine-tuning. This article explains those 3 concepts . #llm #generativeia #ia #rag #finetuning #promptengineering

5

Does AI kill (Agile) programming? Quite in contrary, if you use it wisely. At least this is the message Llewellyn Falco provided in a fascinating meetup yesterday. Llewllyn started where every Agile programmer should start: With the tests. According to Kent Beck *) tests address four needs: - Specification (& Documentation) - Confirmation - Feedback - Regression So Llewllyn fed unit tests into ChatGPT and used some prompt engineering to make it generate running code. He could even make ChatGPT run full test - code - test cycles to mitigate programming errors ChatGPT hallucinated. In a few minutes he was able to generate code, that would have needed five or ten times more time to write by hand. He also discussed limitations of the approach: - ChatGPT is only able to write code with a limited nesting level. There was some discussion, whether this is just a current limitation or a fundamental problem of LLMs. Time will tell us. - The design is... well. "I don't care" was Llewllyn's surprising comment "the code is generated anyhow. If there is a bug, I'll feed a new test case into chatGPT and it will fix the bug". Kicking refactoring out of the TDD-Loop is a pretty bold step. With more experience we will see, whether it is too bold. - Nevertheless, Llewllyn also demonstrated how to use chatGPT for refactoring. "Don't let it do the refactoring itself", he advised "your browser will be way more accurate". But you can use chatGPT to make suggestions for simple refactorings, such as renaming. He obfuscated a piece of code and then asked chatGPT to make at least two suggestions for each variable name. Most of the suggestions were even better than the original piece of code. My personal conclusion is that using LLMs for Agile development is still in an experimental state, but offers fascinating new options for a completely new approach to TDD, which might also extend its usage dramatically. How long it will take until the tool programmers integrate these approaches beyond what Co-Pilot offers, will be crucial for how fast this will come. I'm not sure I buy into Llewllyn's prediction that this will already be true in 2025, but the Agile programming we will teach in five years probably will have very little resemblance with how we work today. Thanks to Eric Lazarus for organizing this great event! More details at https://lnkd.in/dDwrxiAc *) Sorry Kent, I had *too much to do* to actually check the sources, but chatGPT contributed it to you and the four items are cool enough, so they could be your's, so I thought, I'd rather *ask for forgiveness*, if the credits are wrong 😉

42

19 Comments

Thinking Elixir News 202: In this week's edition, we dive into the exciting release of ElixirLS 0.21.0, enhancing the developer experience with new code actions and more efficient dialyzing on the latest OTP. We also discuss José Valim's insightful commentary on Elixir’s upcoming type system, addressing bug-prone comparison operations, and additional advancements in exception handling. Don't miss the unveiling of "Bloom," an opinionated extension to Phoenix core components, alongside Chris McCord's demo of lightning-fast hot code deploys across a global Fly.io cluster. We round off with the legal tussle over the FTC's ruling on non-compete clauses and AWS's S3 billing adjustments that provide relief from unauthorized access charges, and more! #elixir #aws #ftc #devux #developerexperience https://lnkd.in/erkm4F9M

19