Has the Thoughtworks tech radar lost it's mind?
Posted by MattDTO@reddit | ExperiencedDevs | View on Reddit | 65 comments
I was reading through the tech radar, and just to give an example, "AI-friendly code design" was the top technique in "assess". Not only that, the amount of buzz words al over the place make it completely lose focus. What problem are we trying to solve? Let's focus on that. You don't need highly scalable AI-ready distributed product-oriented thinking for industry-standard agent-powered language graphs as a self-hosted PaaS. You need to write some code to solve a problem at your company. I'm just saying, stop adopting and start coding.
Adorable-Fault-5116@reddit
I think it's worth resetting whatever reputation you have formed in your head when you think of Thoughtworks.
Many, many years ago now (early 2000s), they were hot shit because they hired a bunch of well known voices in Java and enterprise software design, at the forefront of Agile, XP, and other good practices. I think it's fair to think it's fair they had that good reputation.
Today they are a washed up body shop, in ability indistinguishable from cheap offshoring, while continuing to squeeze money out of their now very out of date reputation.
TheNewOP@reddit
Never knew this was the case. As someone who graduated a decade or so after that, they were always Revature-tier in my mind
WhiskyStandard@reddit
Oh yeah, I worked with a bunch at an F100 in the early ‘10s. The senior people had some good war stories and were a degree or two from your Martin Fowler types, but toward the end I definitely started feeling a sense that they were training up a bunch of Java people to become Rails people way too quick. A lot of dogmatic adherence to TDD and Scrum ceremonies without caring about the outcomes.
I stuck around years after they left and it wasn’t until I gained the maturity to decouple my identity from Rails that I started asking “wait, why did we use Rails on a read-mostly, high cost/low subscriber premium information site instead of modifying an off the shelf (probably PHP) CMS?” The managers who had been around back then were like “well Thoughtworks convinced the senior leadership [now gone to bigger C-level roles elsewhere] that Rails was the new hot shit and PHP was dead…”
jenkinsleroi@reddit
I mean it basically was hot shit, and then laravel stole all its ideas. And php was and still is terrible.
WhiskyStandard@reddit
I was a PHP refugee to Rails so, yes, I know how bad it was and how much promise of better things Rails brought with it. But after years of working on the codebase and seeing what it had become, what we were able to deliver off of it, and how hard/expensive it was to hire talent, I was faced with the conclusion that although we were clearly going to be a content based site from the get go, no one at this firm with wide ranging industry experience had ever seriously considered any of the incumbent content management systems available at the time.
That probably would’ve been Drupal or Wordpress. Maybe something enterprisey in Java that I didn’t know about as well. And given that most of the TW engineers I worked with had way more Java experience, that’s especially bad.
Of course, the irony is had they done that I never would’ve been working there to eventually come to that conclusion.
jenkinsleroi@reddit
I suspect that the issue was your corporation and not Thoughtworks. Rails would not have been an unreasonable choice, and halfway between a cms and Java.
Adorable-Fault-5116@reddit
Yeah, there was a time where they were at the forefront of doing software development correctly, at least compared to the enterprise java outfits they were talking to (and eg I worked in).
Herve-M@reddit
There was a time where they had great inside & view, but then they lost 90% of the good one in the last decade, limiting to Martin F. and Birgitta B.
Independent-Fun815@reddit
This is not true for many brands these days? For whatever reason, bought out, old heads leave, etc, corporate nonsense, ppl for some reason associate with the name even though it's a shell of the past.
jmking@reddit
I've got over 23 years experience in this industry and I have NO IDEA WTF you're talking about. Context?
jenkinsleroi@reddit
You are out in left field if you've never heard of them or wondered why anyone cared.
That's also not how someone with 23 years of experience and working as a tech lead responds.
jmking@reddit
If you say so
jenkinsleroi@reddit
Thank you for proving your ignorance.
People are talking about the history for a thing over the past two decades, and you claim to have over twenty years of experience and never heard of it, then choose to post a trend search for the last year?
Lame.
jmking@reddit
You can click the dropdown and see the trends from 2004 to present. Nothing changes. Thoughtworks isn't even a blip.
I don't know why the fact I've never heard of whoever these guys are or why they're relevant upsets you so much. How is the fact you've heard of them any more valid than mine having not heard of them?
My account history on Reddit is ancient and this is the first time I've ever heard of this company mentioned here.
So far I've presented quantitative data compared to your "trust me bro". I dunno what to tell you, man. Do you work for them or something?
jenkinsleroi@reddit
You do realize you can't compare search terms across domains meaningfully, bro? Pwc and Accenture are known more for general consulting and accounting than technology leadership. Might as well add Game of Thrones and Taylor Swift on there.
I looked in your history and you were asking questions 11 years ago that were answered 20 years ago by a well known principal at thoughtworks. I hope Reddit isn't your primary source of information.
The reason I commented is because you have a dismissive and shitty attitude that is exhausting to deal with IRL. It's the kind of thing I see from junior engineers who think they know everything. That's why you have negative votes.
In a thread where there are multiple experienced devs discussing a topic, instead of thinking "hey maybe I should check this out" you just said "never heard of it, must be pointless."
20+ years of experience is bs, unless you're counting hobby projects and it help desk.
jmking@reddit
Why didn't you link it? That'd be neat to see.
Wow, jeeze, again - why does the fact I haven't heard of some company so offensive? Do you really think every person in tech knows that this rando company posts something called a tech radar? The post was made with zero context. They didn't even link to wtf they were even talking about.
You're ascribing a tone that wasn't there. I was basically like, ok, who and what and why is it relevant? Still no one has answered those questions. I still don't know why this matters. All I'm getting are indignant, self-aggrandizing digs that fail to even attempt to address the actual topic.
...kinda like a "junior dev who thinks they know everything" would do. hurr hurr
jenkinsleroi@reddit
You are like someone who says they are a professional physicist but has never heard of quantum mechanics. Then when it comes up, says it must not be important because they've never heard of it. Then complains that nobody will explain it to them.
Take the effort to look up thoughtworks and the people named in this conversation. But you probably won't because you haven't already and are too lazy to do it.
And that tone was there. That's why you have negative votes.
tsingy@reddit
That’s what it means for assess right?
tr14l@reddit
Yeah, thought works has those ratings for a reason.
ExperiencedDevs-ModTeam@reddit
Rule 9: No Low Effort Posts, Excessive Venting, or Bragging.
Using this subreddit to crowd source answers to something that isn't really contributing to the spirit of this subreddit is forbidden at moderator's discretion. This includes posts that are mostly focused around venting or bragging; both of these types of posts are difficult to moderate and don't contribute much to the subreddit.
eurasian@reddit
I think they mean "code that is easier for LLMs to parse". That seems a reasonable thing to assess.
Code that's easier for an llm to parse is also easier for a human to parse. And of course, easier for someone new, to use Claude or whatever and get a good picture on wtf this repo does.
Hot-Profession4091@reddit
You’re not wrong, but also not entirely right. A few years ago I was on a team evaluating how to integrate AI into our workflows and we found that the models (at the time) struggled with some patterns that were easy for us, but they were, let’s call them, “nonstandard”. When we changed those patterns to conform more to the training data, what the model “expected”, it performed much better. What we found was that the more we followed community norms, even if those norms are objectively bad, we got better performance from the AI intern.
jenkinsleroi@reddit
This is the right take. AI will often give you something bad because it's been trained on entry-level material and tutorials. Or produce something that's overengineered because it's trying to follow enterprise patterns.
At the same time, if you are doing non-standard things for no reason, this is a good reason to stop being weird and do things the normal way.
Hot-Profession4091@reddit
I don’t know if it’s the “right” take, but it’s a true one.
MattDTO@reddit (OP)
So having a better organized repository is suddenly called AI-friendly design? I think it's just fluff. And if we are going to make a change, we at least need to define how to measure it. For example, maybe you want a "better summary of the repo" or to "make it easier for AI to make changes across the codebase". But ultimately, there is no autonomous coding agent that actually works, so I don't see why we should be "assessing" a technique that hasn't even been established yet.
new2bay@reddit
It’s also about smaller repos. Due to their fixed context window sizes, LLMs don’t do well at all with huge monorepos. You can have a well organized monorepo, or one that’s a ball of mud. Smaller repos have less of a tendency to turn into balls of mud, and they’re also conducive to service-based architectures.
PedanticProgarmer@reddit
If you want to sell the hexagonal architecture, the AI argument works quite well for convincing less technical people.
If the hexagon contains the business logic, the whole business logic, and nothing but the business logic, then it’s the ideal input for LLM that needs to understand what the system actually does.
CalligrapherFit6774@reddit
Sometimes framing it as helping AI can get more people on board with organising the repo well.
sciencewarrior@reddit
How do you measure good code? How do you express the expected value of a refactoring in dollars and time to break even? These are not easy questions, AI or not.
Ninja-Penguin@reddit
I still don’t trust agents yet for writing code but I do use AI a lot for autocomplete and writing isolated functions.
Let’s say I have a chat app and want a simple function that takes in an organizationID and I want all of the images uploaded by its users.
There’s already a function called getUserImages(userId) and a function called getUsers(orgId).
If those functions are clearly named and documented, and the data model is documented in code, then there’s a high chance that just by starting to write
func getAllUserImages(orgId)
it will do the rest successfully.
codemuncher@reddit
Remember when ai was so smart that it would be able to handle any code base and also put every dev out of a job?
And now we are begging people to make the code bases legible to the LLMs.
How far we’ve fallen.
Norphesius@reddit
The promises have all been completely scaled back. The major AI companies have gone from "we'll put everyone out of work and be able to cure cancer" to "its for creating a fake, endless social media stream". Somehow though Zuck still thinks AGI is 3-5 years away.
Its all a big joke, and the punchline will be the the stock market imploding when the bubble finally pops.
sciencewarrior@reddit
This is it. Making the codebase AI-friendly isn't to the detriment of humans. Sensible naming, clear responsibilities, all those things we've known as good practices are being repackaged as "making your code AI-ready".
Ok-Yogurt2360@reddit
That's true untill it isn't. When that point arrives you better be focussing on humans instead of AI.
mothzilla@reddit
Feels like tail wagging the dog though.
Ok-Yogurt2360@reddit
I think you are putting the cart before the horse here.
Abject-Kitchen3198@reddit
So, basically it has nothing to do with AI. Same like saying "write a code that's easy to test so that AI can write the tests".
malln1nja@reddit
I don’t care what it’s called if it finally gets people about the code they write beyond “it works” (which is also not entirely true).
Abject-Kitchen3198@reddit
It's been a struggle since forever. We never lacked people talking, writing and creating tools around all the proposed remedies for it.
blackize@reddit
You are correct about what they mean but I think a bit off on what this means in practice.
This isn’t necessarily “clean code” type stuff but rather colocalization of context. Imagine a web app comprised of 3000 line files that contain a user model, user controller, user serializer, user service, etc. everything relating to user CRUD in a single file. This is the type of thing people are experimenting with.
waltz@reddit
Kinda yeah, but also it's always been a weird laggard. I started subscribing to Tech Radar because some of our clients' leadership teams were soaking it up, and I didn't understand why these execs were so half-informed. I think it's helpful reporting for this kind of person, it helps non-engineering orgs get a snapshot of things that they should be tracking for internal use.
If you work at a company that is engineering-centric all of their advice seems very late and half-too-conservative. Your company probably has a CTO, there are people with the Principal Engineer title who have actual sway on what gets built, eng. teams are a profit center. These kinds of places are already well connected and can do tech evaluations on their own. These shops aren't buying consulting services, they have it onboard. In short, the target audience has changed over time.
jenkinsleroi@reddit
It's useful if you're at a small org in technology, and those roles aren't well defined. Keeping tabs on things happening in multiple fields is hard.
tr14l@reddit
This is something we've been researching. Basically, how would code need to be structure to let a day 1 mid-level with good documentation do most of the work.
To be honest, were assessing, but I suspect a sharp diminishing point that will not make it worthwhile
yetiflask@reddit
1999 called. Bro, thoughtworks is well past its prime. I haven't heard this name in 2 decades.
Cahnis@reddit
Those buzzwords are not for you. The VCs read that and do ahegao faces.
Gr1pp717@reddit
That should be written on the walls of every tech company.
99PercentApe@reddit
They used to make it easy to use their tech radar to create your own by just uploading a CSV. My team made good use of this and it was an engaging way to show what tech we were retiring or investing in. For some reason they stopped allowing this and expected you to build and host the thing yourself. No idea why but it’s always sad when something cool is allowed to just die for no good reason. And yes, their own radar is predictably buzzword dense and irrelevant. Almost like it’s written by Consultants.
flavius-as@reddit
They're utterly wrong, yes.
Have they lost their mind? No.
They produce materials which companies want to read and making money.
They are rational about what makes them money and they do just that. Let it sink in.
Roadhause@reddit
Are we supposed to be surprised that AI tooling and techniques are making the list in 2025? The "assess" category is specifically for things to follow but not necessarily trial at this time. IMO it would be irresponsible to not follow these tools and techniques as a technical leader.
If these techniques improve and actually become viable and sustainable, great, you've got some context to leverage. If they fail miserably, also great, you understand them enough to explain why they won't work to overzealous stakeholders.
deadwisdom@reddit
Yeah, honestly by much of the industry “Assess” is super conservative. The category is “Let’s decide if this is good or not” the assessment can easily be “Nah.”
Jestar342@reddit
"Assess" means literally that. What's the problem?
They are taking a trend and assessing it.
Tomicoatl@reddit
AI is only going to get better and be used more no matter how much you and others in this subreddit fight it. Thoughtworks themselves has been a slop shop for a while however.
Foreign_Addition2844@reddit
Didn't you hear? You need LLM integrations for crud screens now.
MattDTO@reddit (OP)
Oh I must have missed this! We already replace all our crud screens with chatbots
forgottenHedgehog@reddit
I've worked at companies where Thoughtworks "consultants" were brought in, they have a few people who have some idea what they are doing, but then they swap them out for a garbage-tier developers from spain or other cheap country.
The radar is just an extension of their marketing.
FreshPrinceOfRivia@reddit
Thoughtworks is well past its hayday, and even at its peak it was a glorified bodyshop. I wouldn't take their tech radar too seriously, or anything else they produce for that matter.
thy_bucket_for_thee@reddit
Yeah, the idea of a "tech radar" is quite useful for organizations, but there's comes across as "manufactured" if you catch my drift. Have still met a lot of great devs that have come out of Thoughtworks, so at least culturally they're able to export something decent unlike Amazon at least.
mattgen88@reddit
I keep bringing this up at my company. Is anyone tracking ROI on this? Crickets
MattDTO@reddit (OP)
So it's easy to track activity, like how many chats people had with an agent or the number lines of code accepted. I guess we would look at Jira to compare velocity with and without AI? Idk
mattgen88@reddit
We've had like 30 models added in the last month. Not sure what we've built with them though. Lots of funny images!
Old-School8916@reddit
wait it's all buzzwords? always has been 👨🚀🔫👨🚀
demosthenesss@reddit
I'm not sure this has ever not lost its mind.
I remember reading their tech radar many years ago feeling similar.
I've never really read it and gone "you know wow this is profound and insightful." Because it's basically a dump of hype all the time.
bwainfweeze@reddit
The best thing about Thoughtworks has always been Martin Fowler.
They also employed Aaron Schwartz and Jim Highsmith.
Those are the three nicest things I can say about them.
justUseAnSvm@reddit
This is what I found: https://www.thoughtworks.com/en-us/radar/techniques
Like half of these are good ideas!
economicwhale@reddit
I still read this thing - it’s been a bit meh but helpful to see where the industry is. There always reasonably conservative with pushing new things and I find that a helpful filter.