You gave us feedback on Saturday. We made Second Brain more useful by Monday.
We launched Second Brain on Saturday, listened closely to the feedback, and got straight back to work.
Today, v2.1 is live.
The biggest request was simple: make Second Brain more useful in everyday life.
So this release helps it capture more of what already matters to you, explain its answers more clearly, and feel easier to use from the moment you open it.
You can now bring in context from your calendar and email, so Second Brain has a better understanding of your commitments, conversations, and ongoing work.
Recall is also more transparent. When a memory surfaces because of a connection, you can now see why it appeared instead of having to trust a black box.
We improved the desktop experience too. Connections are easier to explore, setup is clearer, and new users get better guidance instead of landing in an empty app with no direction.
We also made recall faster and more focused, so one large result does not overwhelm your AI assistant with unnecessary context.
A lot of work went into making the product more reliable and easier to build on from here. Huge credit to Vincenzo Fabiano, @FaberVi, who made the Worker code modular and much easier to maintain.
This release is not us saying every piece of launch feedback is solved.
It is us showing that when people take the time to tell us what is confusing, missing, or fragile, we listen and act quickly.
The remaining feedback from Saturday is being tracked openly here:
https://github.com/rahilp/second-brain-cloudflare/issues
Thank you to everyone who tried the launch, reported issues, challenged assumptions, and helped shape what came next.
Replies
'when a memory surfaces because of a connection, you can now see why it appeared instead of having to trust a black box' is the sentence that matters most in this release. showing the why is what turns AI output from a suggestion into a piece of evidence. every AI product that goes long-term serious about trust ends up here. its a receipt for the reasoning, not just the answer. weekend-to-monday turnaround on the transparency ask is a signal by itself. proves the feedback loop is real, not marketing. congrats on v2.1.
Second Brain for AI
@thenameisarian Thank you! Receipts for the reasoning is the right way to put it.
Worth saying the limit out loud though, since others landed on it in this thread: the receipt proves how a memory got retrieved, not whether it is still true. A perfectly explained connection can be built on something six months out of date. That is the next thing to solve.
the staleness point above is worth pushing further too - the risky case isn't even one stale memory, it's two stale-but-plausible ones getting connected into a synthesis that's wrong in a way neither source alone would be. an old calendar entry plus an old email can combine into a confident-sounding "fact" that never existed on its own. the why-it-appeared view would look totally clean in that case, since each piece really did connect logically, the problem is just that both inputs expired quietly.
Second Brain for AI
@omri_ben_shoham1 You are right that the why-it-appeared view would look spotless. It explains retrieval, not truth. Worse for your case specifically: contradiction detection here is write-triggered. Two aged entries combining at read time never trip it, because nothing new was written to contradict.
There is an open issue (#228) for a time-triggered pass that flags long-untouched entries and puts an age qualifier on recall. That helps the single-stale-memory case. Your compound case is harder, since both pieces stay individually plausible. Adding your framing to that issue, because it changes what the fix needs to do.
the why it appeared bit is the one i would have picked too. one thing it does not cover though. why it surfaced explains retrieval, not whether it is still true. a memory can be perfectly connected and six months out of date, and the connection view looks identical either way.
calendar and email probably make that heavier rather than lighter. both are full of things that were true when they were written, the meeting that moved, the address that changed, and nobody goes back to correct any of it.
does the recall view show when a memory was written, or only why it came up?
Second Brain for AI
@jernej_jan_kocica Good question, and the answer is yes on the first half.
Every memory stores its written date and that date travels with recall, not just the dashboard. The dashboard shows source, tool, timestamp and status per entry, all editable.
You are right that this does not tell you if it is still true. The current design handles staleness through status and contradiction detection rather than by fading old memories out. The catch you are pointing at is real: contradiction only fires when something new gets written, and entries marked canonical sit on a recency floor, so one that quietly went out of date never surfaces as suspect.
That gap is open as GitHub Issue #228, a time-triggered pass to flag long-untouched entries and add an age qualifier at recall.
@rahilpirani that is more than i expected, and 228 is the right shape.
one thing i would watch on the age qualifier. age and staleness are not the same. some memories never expire, a name, a birthday, where someone grew up. flag those by age and the qualifier turns into noise people learn to skip. the ones worth flagging are the memories about something that can move underneath you, status, address, plan, which tool they use.
cheap version might be whether the memory describes a state or a fact. good luck with it, will watch that issue.
I appreciate that you shared the remaining issues publicly instead of pretending everything is finished. That kind of transparency usually builds more trust than a perfect looking roadmap.
Second Brain for AI
The open issue tracker is a strong way to turn launch feedback into visible priorities. Curious how you decide what ships immediately versus what stays open for deeper design work?