Gartner has been out of touch with tech analysis: and their latest report on AI coding assistants show just by how much. - They rank Amazon, GitLab, GCP, Windsurf all above Cursor - No mention of Claude Code or OpenAI Codex Cursor pays Gartner nothing: others do Embarrassing

Don't want to say I called it, but... you know... sometimes I know what I'm talking about. www.techdirt.com/2024/11/27/b...

Beth certainly will!

NEW POST I'm about to head off on travels for the next few weeks. Before I go, here's a bunch of scattered thoughts about Gen-AI and software development martinfowler.com/articles/202...

NEW POST Legacy Modernization is hard, especially when we've lost the source code. Thiyagu Palanisamy and Chandirasekar Thiagarajan describe how AI helped them reverse engineer. martinfowler.com/articles/bla...

NEW POST The Bahmni open-source hospital management system was built with an AngularJS front-end over OpenMRS REST API. Rahul Ramesh is using AI tools to convert it to React over FHIR. martinfowler.com/articles/res...

NEW POST CLI coding agents are powerful tools but don't understand your particular environment. Ben O'Mahony shows how to assemble your own. martinfowler.com/articles/bui...

NEW POST Unmesh Joshi and I had an interesting email conversation about how when programming with an LLM he likes to grow a language of abstractions. martinfowler.com/articles/con...

NEW POST When I used to give talks, I was very good at keeping them to the allotted time, even with a last-minute change. Key tool: Expansion Joints. martinfowler.com/bliki/Expans...

NEW POST OKRs are popular but often set in a top-down cascade. Paulo Caroli describes a path for teams to create their own, in a collaborative process to align with broader strategy. martinfowler.com/articles/tea...

FINAL § @srny.bsky.social concludes his article in impact intelligence by addressing five common objections to this activity, including slowing down, lack of agility and collaboration, and the unpredictability of innovation. martinfowler.com/articles/imp...

This is why the whole thing is so hard to trust and if you have any critical thinking, you stay away from such rankings. The same is true for Gartner, where many vendors also pay very heavy sums to Gartner to have Gartner evaulate them

So to be impressed by a machine going way beyond the question asked feels like the technical implementation of this exact anti-pattern. Worse, all that generated output creates significant gravity/path dependence, your brain turns off all that much faster. Why do we assume this is good?

Five Quick Reads: Durable v Disposable code, Domain Models as memory, getting AI to generate its own coding rules, what AI should stand for, and what loses more time than you save with AI martinfowler.com/articles/202...

NEW § @srny.bsky.social continues his article on impact intelligence by outlining five actions that can be done to improve impact intelligence including: demand management, impact validation, and an alternative to ROI. martinfowler.com/articles/imp...

Nik Malykhin found it useful to get an AI assistant to write its own coding rules by analyzing his code, and then asking it to refine them as worked with it. nik1379616.substack.com/p/can-we-mak...

NEW POST Sriram Narayan looks at how to properly assess the effectiveness of software development - by developing an intelligence about its business impact martinfowler.com/articles/imp...

We recently ran an experiment to explore how far GenAI can currently be pushed toward autonomously developing high-quality, up-to-date software without human intervention, and gather observations about where it breaks down. martinfowler.com/articles/pus...

My pulse will be quickenin'

NEW POST @xpmatteo.bsky.social shows why the common metric of AI code acceptance has big hole. LLM can be helpful even if you throw away its code. martinfowler.com/articles/exp...

NEW POST Many people like to give their LLM assistant a persona, such as a stubborn donkey, a genie, a slot machine, and Uriah Heep martinfowler.com/articles/who...

Korny Sietsma has a great example of how using an LLM for coding is very helpful but with limitations. blog.korny.info/2025/07/18/a...

Thoughtful essay on value and limits of AI. "One of the real challenges here is the way that A.I. undermines the human value of attention, and the individuality that flows from that." www.nytimes.com/2025/07/18/o... (gift link)

I've known Beth for ages: wicked smart, understands tech really well, and cares about its impact on ordinary people. We need people in Congress like that.

Isn't this a perfect evolution of form and function? #Paddington #nonslip

Being afraid of failure doesn’t help people succeed. It helps them stay safe by succeeding less often. When we make failure safer & cheaper, we can succeed more.

I really resonated with: "I think I’m less interested in my own happiness (whatever that means) than I am interested in doing work that feels worth doing" - @charity.wtf reflects on her career charity.wtf/2025/07/09/t...

Pete Hodgson continues his excellent writing on Gen AI by proposing an approach to leading engineers towards an AI-assisted future, centered around a the concept of aligned autonomy. blog.thepete.net/blog/2025/06... @thepete.net

Gen AI breaks two common constants with developer tools: developers are willing to flit between Gen AI tools and they are willing to pay for them. Implies it's not too late for new tools , enterprise adoption will be slowed by a lack of consensus redmonk.com/sogrady/2025... @sogrady.org

Makes me realize how hiring is not focused on this stuff: "how would this person make the team better." I guess, it is hard to be focused on this. But this was one of the *very* rare devs who made every team much better. Nicer. More motivated. More a "team." Still think of it

NEW POST Even with LLMs, @birgitta410.bsky.social still cares about the code: “LLMs are NOT compilers, interpreters, transpilers or assemblers of natural language, they are inferrers. martinfowler.com/articles/exp...

Informative interview about a company that's leaned heavily into AI Augmented software development. Interesting to hear that this is leading them to increase the hiring of grads.

FINAL § In complex environments, the characteristics of Expert Generalists lead them to be particularly valuable in driving tasks to completion. Unmesh, Gitanjali, and I thus complete our article by summarizing the value of this skill. martinfowler.com/articles/exp...

NEW § While we've spent this article praising the Expert Generalist, Unmesh, Gitanjali, and I simultaneously do not deny the value of specialist knowledge. We also note that EGs work well with LLMs. martinfowler.com/articles/exp...

NEW § To grow Expert Generalists we need to focus attention on fundamentals rather tools. As an example, Unmesh, Gitanjali, and I describe a workshop we've used to break silos of application development, data engineering, and devops martinfowler.com/articles/exp...

NEW POST Gen AI for software development is the biggest jump up in abstraction since assembler to HLLs, but also a jump sideways. martinfowler.com/articles/202...

NEW § We have two crucial checkpoints for spotting — and then nurturing — expert generalists. Unmesh, Gitanjali, and I look at hiring and career progression. martinfowler.com/articles/exp...

I wrote a guest post on The Pragmatic Engineer newsletter.pragmaticengineer.com/p/two-years-... rounding up 2 years of using AI coding assistants - how they evolved; ways of working; impact I see on speed, quality and team flow; and some thoughts on the future

NEW § Three more characteristics for Expert Generalists. Unmesh, Gitanjali and I describe favoring fundamental knowledge, skills blend, and sympathy for related domains martinfowler.com/articles/exp...

NEW POST Increasingly our industry favors deep specialists. But we've found our most effective colleagues have a broad-based skill that's based on spanning many specialties. Unmesh, Gitanjali, and I write about explicitly recognizing this as a first-class skill martinfowler.com/articles/exp...

From one of the people I trust most on LLMs - a clear explanation of a common and very dangerous security threat

This is a great interview. Kent Beck talks to @gergely.pragmaticengineer.com about how he's having more fun programming now with AI genies than he's had in ages. Also: the birth of Extreme Programming. newsletter.pragmaticengineer.com/p/tdd-ai-age...

"We always assumed there would be people gunning for it, but we also knew it was something people wanted. We knew it was an idea whose time had come." — @merici.bsky.social

Birgitta guides us behind the scenes of an "autonomous coding agent"

I'm wondering what to do about analytics on my site. Difficulties getting Google to work with less tracking, and alternatives not worthwhile. martinfowler.com/articles/202...

worthwhile🧵on the remote-v-office-v-hybrid topic

NEW POST Possibly my favorite writing tip: read your draft out loud. martinfowler.com/bliki/SayYou...

Two legends, one epic convo! 🔥 ‪@martinfowler.com‬ and @boicy.bovon.org‬ dive into the early days of Agile and ask the bold question: Is Design Dead? Don’t miss this insightful chat! youtu.be/xsMUuOwv7IA?...

Once again, if your LLM system combines access to private data, exposure to malicious instructions and the ability to exfiltrate information (through tool use or through rendering links and images) you have a nasty security hole This time, GitLab: simonwillison.net/2025/May/23/...

Whenever I see 3 significant digits in a benchmark, especially of a qualitative task, I immediately get suspicious. "Sonnet 4 improved our SWE-bench agent single pass score from 60.6% to 70.6%" So, 60 to 70? What kind of errors does it still make? What new errors does it make?