Where Pensiv is heading.
What works today, what we're building next, and the honest difference between the two. No roadmap dates we can't keep.
We believe in telling you exactly what you're getting. Not the version we're planning to ship. Not the version that exists in a spec somewhere. The version that's running today — and separately, the version we're working toward.
Here is where we are.
What works today
Five things are real and working right now.
Connects to the tools you already use. Calendar, email, Slack, GitHub, Linear, your drive, and more — fifteen tools today, and growing. Connect one once, and new work flows in on its own. You can also ask Pensiv questions in a thread that remembers the last answer, so a follow-up builds on what came before.
Knowledge that carries across sessions and models. Every question your team asks, every judgment call it makes, every document it works through — that context is stored and searchable. Not just within a session. Across sessions, across teammates, and across whatever AI model you decide to use. Switch models next month. Your team's knowledge comes with you.
Runs anywhere, including offline. Pensiv runs on your own hardware. It does not need a cloud connection to work. This matters if you operate in environments where data leaving the building is not an option. It also matters if you just want to know that what your team has figured out stays yours.
A source trail on every answer. Every claim Pensiv surfaces traces back to something your team actually said or documented. You can follow that trail. An outside auditor can follow it. There are no answers that arrive from nowhere.
Pays back fast. An independent economic audit found Pensiv pays back in months for teams that depend on accumulated expertise. The median deployment clears itself in under three months. We published that audit in full — read it here.
Important knowledge persists. Less important knowledge fades. The system forgets the way a good colleague forgets — gradually, on its own, without you having to manage it.
What we're building next
Here is what is not done yet. We will tell you when it is.
Knowledge that reorganizes itself.Today, Pensiv stores and retrieves. What we're building is knowledge that quietly reorganizes itself — connecting things your team learned six months ago to what it's working on now, so that tomorrow's answers are sharper than today's without you having to do anything differently. Think of it as the work that happens overnight, so the morning briefing is better than the one before it.
We check an insight before we trust it.Not everything a team writes down turns out to be right. We're building a step that checks candidate answers against real outcomes before Pensiv promotes them — so the knowledge that rises to the top is the knowledge that held up, not just the knowledge that sounded good at the time.
Measuring whether the system is getting better at your work. Public benchmarks measure how AI performs on standardized tests. We're building something different: private measurement tied to your team's actual outcomes, so you can tell whether Pensiv is getting sharper at the specific work you do — not a generic evaluation that doesn't reflect your domain. When this is ready, you will not need to take our word for it. You will have your own numbers.
How we think about honesty
AI products are full of roadmaps dressed up as current capabilities. We find that disrespectful to the people evaluating them.
If something is built, we say it's built. If something is being built, we say we're building it. If something is an idea we're thinking through, we do not put it on the website.
This is not modesty. It is respect for the fact that you need to make real decisions — about your team, your data, your tools — and you cannot make them well if the vendor is telling you what they hope will be true instead of what is true right now.
One more thing
We are a small team. We build deliberately. We ship things that work rather than announcing things that don't exist yet. If you want to know when the things above are ready — or if you want to shape what we prioritize — the waitlist is the right place to be. Early access customers inform what we build.
What comes after
The long arc is this: every hard-won thing your team learns should compound. Session after session, model after model, team member after team member. The knowledge a senior person builds over years should be available to the whole organization — not locked in one person's head, not lost when they leave, not invisible to the AI tools you bring alongside them.
The knowledge infrastructure to do that is what we're building. Some of it exists. The rest is on the way. We'll keep telling you which is which.