I've been fortunate enough to be given the opportunity to try Devin Desktop for free over the past ~month. I tried it while using SWE-2 exclusively, as that is what really makes it different (Using Opus 5.5 on Devin vs using Opus 5.5 on Claude Code vs using Opus 5.5 on Cursor would not really show much difference).
SWE-2 is very interesting to build with. It can be insightful, provide highly technical answers and very thorough reviews of code ...
... but it can also be rigid, miss things that are easy to catch, and outright ignore instructions. I told it to convert a JSON list to a DB table & entries and to remove 100% of any and all reliance on the JSON list ... and I had to repeat my instruction 5 times before reliance on the list was actually removed. The project I was working on is quite large ... but I kind of felt like a direct command should have been followed better than that.
Best I can tell, it was VERY reluctant to remove reliance on the list because the project has an existing production deployment and the change would disrupt existing saved entities. But we had dealt with those and walked through processes where we backfilled for existing entities, saved our processes against needing the list, etc. It just couldn't grok the whole thing at once ... and so kept falling back to being "safe".
Again, though, SWE-2 can be very insightful. It gave great code-reviews on PRs submitted by other developers. It caught my Grok 4.6 being sloppy in some places. It explained thoroughly things that were wrong or missing or not properly secured.
I found SWE-2 to be MUCH better at review and analysis of existing code than at writing new code. It would provide good recommendations, but if I fed those recommendations to another AI, that other AI (Opus 5.5, Grok 4.6, even Composer 2.5) would do a better job implementing them than SWE-2 did.
Overall, I'm coming away from my month with Devin with a positive impression of it, but with my ultimate workflow of using Cursor as my preferred IDE unchanged. I will, however, DEFINITELY keep an eye on Devin and I anticipate that it will be a STRONG contender. I may even spring for a low-end monthly subscription and keep using it in parallel just for its code review capabilities.
A HUGE win for one of my projects:
We just landed The Cleveland Project for ExtraRep.com
Roughly 10,000 kids and their coaches playing basketball. Extra Rep is helping them manage their organization, keep kids in contact with coaches even during the off-season, provide training, etc.
I'm not just excited by the numbers, I'm also excited because it's a charity that uses sports to improve lives. We couldn't have landed a better first major customer.
Our weekly Space is starting in ~20 minutes! Come hang out and chat about tech, your projects, and whatever else comes up!
xitter.cf/i/spaces/1NxaroLWMPbKj
I may have been laid out recently, but I have heard of the @eMudhra incident. It appears their corporate theme song has been on YouTube for years already:
youtube.com/watch?v=u8ccGjar…
Anyone else feel that actual phone calls are now entirely, 100%, useless? I get so many attempted spam calls that I put my phone on permanent do-not-disturb a year ago. Only people on my contact list can ring me. And they don't -- they use Telegram, Discord, WhatsApp, and Line.
Also ... I'm still completely laid out from whatever bug I caught. This one hit me hard. I'm sleeping in 1-2 hr fits, having fever swings, all kinds of fun.
Got hit by something yesterday ... fever, sinuses, all the super fun stuff. Feel like I need to just lay down and sleep all day.
Of course, that's not possible. But at least I have Bots and AI to help me struggle through.
Going to try to find another host for our weekly Space.
Martin hits the perfect advice for anyone entering the game we're all playing:
> Make the game infinitely playable.
That's it. That's the goal. Keep playing the game. You have to be on the field to win, so make sure you are set up to keep playing.
Also, langmaxxing.com is legitimately awesome. I'm already an addict and it's only been a few days.
My Grok Bot just got an email address. Now I can have an automated customer service experience just as annoying and useless as Enterprise companies 😅
Actually, it's a really cool idea. A way for one person's bots to interface with another person's bots. Anyone up for an experimental project where we have our two teams of bots communicate, split work, and build together?
I'm currently running both Cursor and Devin as I dev each day. I'm running a comparison between the two IDEs and how well they work for me, my workflows, and on my code.
Devin is running SWE-2 High
Cursor is running Grok 4.6
It's still early, but I've already seen each grab a win from the other.
SWE-2 is good at being very thorough in investigations ... sometimes to its own detriment. It overcomplicated a fix for a bug I asked it to identify ... but it ALSO gave me a very good and thorough review of an app's performance and potential improvements that Cursor Grok 4.6 missed.
Cursor is the IDE I'm familiar with and have been using for a long time ... but I'm always open to other options and Devin is doing a decent job so far.
I've got mixed feelings about the X Subscription.
1 - I'm a Cursor Ultra subscriber ($200 / mo)
2 - I'm a Premium+ X subscriber ($345 / yr)
You'd think I'd be very happy to save $345/yr and have everything linked together. A couple of things give me the mixed feelings:
- I use Grok via my X subscription quite a bit for random discussions and explorations (the latest was about the Yamnaya migration from the pontic steppe into western Europe). If queries and explorations like these are going to start counting against my Cursor limits, I'm going to start thinking twice before engaging in them. Even if it'd be fine.
- One of the reasons I use Cursor is because it has always given me access to multiple different frontier models, wrapped within one subscription. I can try out Opus 5.5, then switch to Grok 4.6, then explore Gemini or Muse ... all WITHIN my subscription. The new text on this subscription is "third-party models remain available in Cursor at API list prices". I don't like that.
- The plans don't currently define what they mean by "usage" ... the definition is that one plan gets a "usage" amount and other plans get x4, x15, x50 that. Well ... what are we x50-ing?
Things that happen during our weekly Spaces stand-up and tech chat meeting ...
@DimitriGilbert had his agents run for 9 days, uninterrupted, to build slopCAD. Yes, a computer aided drafting tool.
Today, I suggested he have an agent interface. He already does. So, I told Grok Bot to connect to slopCAD and design a coffee cup ... and all of us just kept chatting while it worked.
Ok, the cup is super basic and the handle definitely would not work, but it's just really fun to hang out with other founders/creators/indies from X and talk and play around together.
We meet every week, and everyone is always welcome!
Replying to @AtSynct
Just had Grok Bot create a coffee cup design using @DimitriGilbert 's slopCAD (slopcad.dbuild.dev)
Today is the day for our weekly stand-up and tech chat. Come join in on the fun. We love meeting new people.
@Aevmorfop @DimitriGilbert @JeremyLasne and I started it as a Discord chat long ago and decided to move it into a public space.
xitter.cf/i/spaces/1MJgNbeAbZkGL
Just had Grok Bot create a coffee cup design using @DimitriGilbert 's slopCAD (slopcad.dbuild.dev)






