First thing I noticed is Bolt, Coolify, Lovable all in adopt, in addition to v0 in assess.
The goal of a tech radar is to map out and try to contain the chaos in a large organization, so you don’t end up with seven teams adopting different technologies, and the evaluation process is more structured. Only one of those should have made it to adoption unless they serve wildly different purposes?
Hi ricardobeat, you just perfectly touched on our CTO's main concern. :) Although we are not a large organization (we are a 3 people team), we have a few technologies around.
Most of that was urged by me the CEO. Some, admittedly motivated by a "red shiny object" syndrome.
We use Coolify to accelerate deployment (atop of Azure), so it does not interfere with Bolt and Lovable.
Our backend developer used Bolt in that past to built frontends for his backends. I use Lovable primarily as an alternative to Figma, i. e. for communication new ideas. None of that ends up in the production product (here we use Cursor mainly).
v0 is in asses because we heard good things about their full-stack capabilities (something that Lovable/Bolt cannot really do).
Here is the summary:
We score each technology across three dimensions: market adoption (how many big companies are really using it and seeing results), relative impact (on our bottom line), associated risks, and internal prios.
For each dimension we have a scorecard along the lines of: if five top 100 startups are using it publicly it gets 3 points of market adoption. The scorecard is far from perfect, but it gives us a good, repeatable algorithm across time and trend.
In the case of GPT-Codex: it is on hold because the perceived relative impact for us is low (we already have a good setup with alternative tools; bolt + cursor). In the above-article I have also linked to our Google Sheet. It contains the scores (some redacted) including the underlying data.
> In the case of GPT-Codex: it is on hold because the perceived relative impact for us is low
Interesting ...
so for you "Hold" ~= "low impact / monitor ", while ThoughtWorks use "Hold" more like "don't start anything new" ( https://www.thoughtworks.com/en-us/insights/blog/technology-... ) [1]. Personally, I've started to read "Hold" in their sense - basically as "not recommended" for new work.
that’s why I first thought "GPT-Codex" (one of my favorite models) was already "not recommended." :-)
[1]
""
Hold: The original intent of the hold ring was "proceed with caution", to represent technologies that were too new to reasonably assess yet. But it has evolved into more of a "don't start anything new with this technology." You may be constrained to use it for existing projects because it is so deeply embedded into the tech portfolio, but you should think twice about using this technology for new development.
"""
"HOLD — Technologies not recommended to be used for new projects. Technologies that we think are not (yet) worth to (further) invest in. HOLD technologies should not be used for new projects, but usually can be continued for existing projects."
it’s my current favorite, with claude-code as runner-up.
As someone with aphantasia[1] I naturally lean toward backend / abstract modeling ( maps, technology radars, databases ) , so non-visual tools like codex-cli, claude-code, or even https://omarchy.org/ are especially appealing.
I haven’t yet seen a technology radar that makes a visual vs. non-visual distinction. If you ever run a survey, adding such a category could be interesting - surprising patterns might emerge.
First thing I noticed is Bolt, Coolify, Lovable all in adopt, in addition to v0 in assess.
The goal of a tech radar is to map out and try to contain the chaos in a large organization, so you don’t end up with seven teams adopting different technologies, and the evaluation process is more structured. Only one of those should have made it to adoption unless they serve wildly different purposes?
Hi ricardobeat, you just perfectly touched on our CTO's main concern. :) Although we are not a large organization (we are a 3 people team), we have a few technologies around.
Most of that was urged by me the CEO. Some, admittedly motivated by a "red shiny object" syndrome.
We use Coolify to accelerate deployment (atop of Azure), so it does not interfere with Bolt and Lovable.
Our backend developer used Bolt in that past to built frontends for his backends. I use Lovable primarily as an alternative to Figma, i. e. for communication new ideas. None of that ends up in the production product (here we use Cursor mainly).
v0 is in asses because we heard good things about their full-stack capabilities (something that Lovable/Bolt cannot really do).
imho - it’s not enough to list trends - the reason behind the status is key.
for example: "GPT-Codex" = Hold. Why?
I find tech radars very interesting - especially in today’s fast-moving world.
A good reference is the ThoughtWorks Radar: https://www.thoughtworks.com/radar
My other favorite: https://www.oreilly.com/radar/topics/radar-trends/
hi pella, thanks for your comment, especially "the reason behind the status is key."
We built the radar primarily to have a data-driven reasoning behind it. I have explained our approach in more detail here: https://blog.tryresearchly.com/articles/built-own-technology...
Here is the summary: We score each technology across three dimensions: market adoption (how many big companies are really using it and seeing results), relative impact (on our bottom line), associated risks, and internal prios.
For each dimension we have a scorecard along the lines of: if five top 100 startups are using it publicly it gets 3 points of market adoption. The scorecard is far from perfect, but it gives us a good, repeatable algorithm across time and trend.
In the case of GPT-Codex: it is on hold because the perceived relative impact for us is low (we already have a good setup with alternative tools; bolt + cursor). In the above-article I have also linked to our Google Sheet. It contains the scores (some redacted) including the underlying data.
Also if you klick on some of the trends (e. g. GEO) it links to our written-out rationale (e. g. https://blog.tryresearchly.com/articles/wie-wir-uns-auf-geo-...)
Thanks for the Oreilly radar. Didn't know this one.
> In the case of GPT-Codex: it is on hold because the perceived relative impact for us is low
Interesting ... so for you "Hold" ~= "low impact / monitor ", while ThoughtWorks use "Hold" more like "don't start anything new" ( https://www.thoughtworks.com/en-us/insights/blog/technology-... ) [1]. Personally, I've started to read "Hold" in their sense - basically as "not recommended" for new work.
that’s why I first thought "GPT-Codex" (one of my favorite models) was already "not recommended." :-)
[1]
Thanks for the links - makes much more sense now!EDIT:
Zalando definition ( https://opensource.zalando.com/tech-radar/ )
exactly. Maybe I need to update our ring-definition or least define them on the radar. Thanks!
BTW: what's your experience with GPT-Codex?
Still in the honeymoon phase with gpt-codex
it’s my current favorite, with claude-code as runner-up.As someone with aphantasia[1] I naturally lean toward backend / abstract modeling ( maps, technology radars, databases ) , so non-visual tools like codex-cli, claude-code, or even https://omarchy.org/ are especially appealing. I haven’t yet seen a technology radar that makes a visual vs. non-visual distinction. If you ever run a survey, adding such a category could be interesting - surprising patterns might emerge.
[1] https://hn.algolia.com/?query=aphantasia
I just put the Chipotle app from "Hold" to "Adopt". Do people really use cringy stuff like this?