Jarvis, six weeks in: what's actually working
Personal · 2026-08-12
Product · 22 April 2026 · 4 min read
Jarvis, six weeks in: what's actually working
Right, so I promised this one a while back and kept putting it off, mostly because I didn't want to write a review before I'd actually broken the thing a few times. Now I have. So here we are.
Jarvis is the tool I've been leaning on for the day job build, the one I mentioned in passing a few posts ago and never properly unpacked. Quick context for anyone new: it's meant to sit somewhere between a project manager and an assistant, chewing through tasks, keeping context across sessions, that sort of thing. Sounds great on paper. Most of these tools do.
I'm not going to do the thing where I list features like it's a spec sheet. That's not useful to anyone, and honestly it's not how I think about tools anyway. What's useful is: does it save me time, does it get in my way, and where does it just quietly fall over.
What's actually good
The context retention is genuinely solid. I can pick up a thread from three days ago, half explain what I need, and it doesn't ask me to re-explain the whole project like some tools do. That alone has saved me a stupid amount of admin time. Small thing, but it adds up.
It's also fast at the boring stuff. Drafting first passes of docs, restructuring messy notes into something coherent, chasing down inconsistencies in a spec. None of this is exciting to talk about but it's the stuff that eats hours if you're doing it solo, which most of us reading this are.
I'll be honest, I didn't expect to actually enjoy using it day to day. But there's a rhythm to it now that feels less like fighting software and more like, I don't know, delegating to someone who's mostly competent.
Where it falls apart
Right, the bit you're actually here for.
It struggles with ambiguity. If I'm vague, it doesn't ask the right follow up questions, it just guesses, and the guess is often confidently wrong. For a tool built around "context", that's a bit of a problem, you know? I've lost time to this more than once, having to unpick a whole thread of decisions because it ran off with an assumption I never actually made.
Longer sessions degrade. I've noticed after maybe forty minutes of back and forth, quality drops, responses get shorter, less considered, occasionally just wrong in small annoying ways. Nothing dramatic. Just enough friction that I start double checking things I shouldn't need to double check.
And the integrations are patchy. Some connect beautifully. Others feel bolted on, like an afterthought someone shipped to tick a box. I won't name which, because it changes, and slagging off a specific integration feels a bit pointless when it might be fixed by the time you read this.
The honest bit
Here's what nobody selling you a tool wants to say out loud: it's not a replacement for actually knowing what you want. If you go in vague, you get vague back, just faster. That's not a Jarvis problem specifically, that's every one of these tools, but I think it's worth saying plainly because the marketing around this stuff never does.
For indie hackers and small teams, the value isn't magic. It's compression. It takes the boring 30% of your day and squashes it down, freeing you up for the bit that actually needs your brain. That's it. That's the whole pitch, once you strip the hype away.
Is it worth paying for? For me, yeah. For someone earlier stage, still figuring out their actual workflow, maybe not yet. Get your process sorted first, then bring in the tooling. Otherwise you're just paying to be confidently wrong faster.
Anyway. That's Jarvis, unfiltered, as promised. More on this as it breaks in new and interesting ways, because it will.