kosovan retweeted
Meet Eney by @MacPaw, a proactive AI assistant for Mac and iPhone.
It flags what actually needs you: the unanswered email, the call you haven’t prepped for, the Downloads folder nobody wants to open. Then it gets the next step ready.
Start with a ready-to-use workflow, or just say what you need.
Skip to done with Eney – macpaw.onelink.me/OlF7/x_org…
We have been working on Eney since 2019.
It is finally live today on Mac and iPhone 🔥
The goal was always software that gets the work done, not one more chat window.
Thank you to everyone who stayed with it!
Meet Eney by @MacPaw, a proactive AI assistant for Mac and iPhone.
It flags what actually needs you: the unanswered email, the call you haven’t prepped for, the Downloads folder nobody wants to open. Then it gets the next step ready.
Start with a ready-to-use workflow, or just say what you need.
Skip to done with Eney – macpaw.onelink.me/OlF7/x_org…
Underneath is partially our own stack. Elix runs inference on Apple silicon, and Mnemos is the memory layer that updates when facts change.
In practice it spots what needs you, prepares the next step, and puts it in front of you to approve. Call starting? Notes, tasks and a follow-up invite are ready when it ends.
Eney is live today on Mac and iPhone: eney.ai
Included with Setapp, with a free trial.
In general I think on-device AI needs this kind of evidence more than cloud AI does, because the result depends on the model, the silicon and the runtime together. Therefore we will keep publishing as we improve. What would you want us to measure next?
Mnemos was measured against three other memory systems on 12 benchmarks with the same conversations, the same answering model and the same judge. Highest on 8 of 12, including 89.0% on LongMemEval-S. Vendor-published scores use different setups, so we did not mix them in.
The number I care about most: decode is 27% faster than the six-engine average overall, and 44% faster at 32K-token prompts. Long context is what an agent with memory actually runs on: your files, your history, the conversation so far.
Elix was tested against six other engines on Apple silicon, identical model weights, 25 chip-and-model combinations. Fastest in 19, within 5% of the leader in 4 more, behind in 2. We kept the 2 on the page.
Proud of our team! 😍
The part I like most is that every score is published as measured, including the runs where we are behind. Same weights, same hardware, same judge for everyone.
We shipped Email Cleanup in @CleanMyMac this week. The question that took longest: how much should software do in one click to years of email? Our answer: group by sender and year, preview everything, keep every delete reversible.
In general I think agents need the same rule.
Most of this week's agent news was about where agents run. The part I find more interesting is the rule for when an agent stops and asks. In Eney the rule today is simple: anything with consequences waits for your confirmation. Where would you draw that line?
kosovan retweeted
Great minds think alike. We just had a head start.
And now, @eneybymacpaw’s preview is open — free, with 20,000 credits included. Our only ask: tell us the truth. What works, what doesn't, and what you wish it did.
Feedback goes straight into the app (Settings → Feedback) or our Discord.
Join in → bit.ly/4hjMrYK
Today the biggest company in AI shipped an always-on assistant with a character that does tasks.
We have been building one for the Mac since 2025. Same conclusion, one different decision: Eney runs on your Mac, and you sign every consequential step. That is the next argument.
Eney in beta for Setapp users, May 2025: 9to5mac.com/2025/05/22/macpa… — and where it is today: macpaw.com/eney
