Humfrid

Nothing beats great user feedback

Today we’re announcing Humfrid. Humfrid is the agent product teams actually want to work with - from first idea to shipped product. Tell him your goals and he works proactively alongside you and your team in Slack to reach them. What does Humfrid actually do? 🐦‍⬛ He keeps track of your goals. Following up on metrics long after you've moved on to the next bet, and noticing when you're off track before anyone else does. 🐦‍⬛ He sharpens your thinking. Surfacing evidence with the numbers and customers quotes you need, drafting the questions you haven't asked yet, and pushing back when you rush to build before you've understood the problem. 🐦‍⬛ He keeps everyone aligned, without the meetings. He catches contradictions the moment they surface - when two teams are solving the same problem, a decision doesn't follow your product principles or when the PRs don't match the brief. Most agents are endlessly capable, but opinion-free. Humfrid isn't - he has a point of view about how product should work. Most agents are agreeable. Humfrid isn't. He's on your side. Meet Humfrid here - it's free to get started. → humfrid.com

Looking for 5 product leaders (CPO / Head of Product / Product Director / Founder) who want AI to proactively work for them and their team by end of Q3. We'll handle: - Context curation - defining what docs, decisions, and signals your LLMs actually need to create relevant outputs - Applied workflows - mapping AI into how your team researches, plans, and ships - In-house sessions - working directly with your PMs to identify where AI multiplies their judgment, not just their output What you'll walk away with: - A system that surfaces what your team needs to reach their product goals before they have to ask for it - PMs spending more time on the high-leverage work - research, strategy, customer decisions - PMs who aren’t overwhelmed by the constant stress of staying on top of things - A product team that treats AI as a multiplayer tool, not a solo shortcut DM me or comment below if this is you.

Agents will be like people. Generic agents will be like interns. Capable of following instructions and doing the work. The real value will come from its context, where preloaded context and memory architecture will be the equivalent of experience and mental framework.

@Steve_Yegge You mean like this

Ubering into downtown San Francisco, every ad looks like desperation. All agent stuff nobody will use. Haven't seen anything that a CIO should or would buy. It's all starting to look like the same ad.

a memory you can add to, edit and curate is more than great ux

obsidian meets jarvis, for product managers

everyone's talking about the importance of context but what context?

/goal as it should be actual business objectives as the goal key results as verifiable outcomes problem spaces, possible solutions and tasks as the work tree

@geoffwoo did

announcement: openai data point i care about: - not model leaderboard screenshots - not “agentic” landing pages - not the 11th copilot for sales managers send me the product where GPT-5.5 turns a weekly human judgment loop into a 4 minute API call with an audit trail.

this is your sign up for momental set a goal, that’s it

For every task they create, they track what context they used If they create new data points, learnings, decisions or even principles, they land in a "memory inbox" for approval And when a human should take it back depends on autonomy setting - we have three: 1) Active: they just do what they're assigned 2) Semi-autonomous: they keep working until they run into something that needs approval (pr merge, sending emails etc) 3) Autonomous: they just keep going (but still adhere to a budget you set on the objective) When they want a human to review, they assign it to a human - they keep track of who should get it, both based on who created the task and what expertise everyone seem to have

@momentalos i like the okr angle. the hard part is making the handoff inspectable: what context did the agent get, what did it decide, and when should a human take it back?

Assign an agent to a key result - they'll take care of it. Break down goals into tasks, based on your analytics, customer conversations and strategy. Drag and drop cloud agents, your human team member or your Claude Code. Agents connected via MCP picks up the task with full context of what needs to be done at why.

I used to make goal visualizations in Miro and then try to capture the why and big picture behind every task in Jira. It was a mess. But humans absorb the implicits and keep a model of how everything fits together. Agents don't. Well, didn't. In @momentalos, agents keep the hierarchy of every task in their context when picking up work, so they know why they're doing it. It's the most important context they can get. And for human 👀, you can seamlessly switch between a view of your goal, and a kanban of tasks.

@Tristanrhee3 Yes! Be kind.

Underrated advice: Become a purveyor of encouragement. Notice what's good in people. Tell them when they do something well. Celebrate their progress. Root for them to win. Be inspired by their success. The world already has enough critics.

@annieqyang clocked in

clocking into work everyday as an indie hacker = posting on X

@robjama 🤐

build in public or build in silence?

The hard part was always knowing what to build. And when you can build or do anything, that choice becomes even more important, and paralyzing.

freshened up our website