Mathias Kleverud

founder @momentalos | shared memory for product orgs | extraordinary alien

[Mathias Kleverud @mathiiias]

/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

[Mathias Kleverud @mathiiias]

@snoopy_dot_jpg we're making our best attempt with @momentalos - it's defining the greater loop (business goal -> metrics to verify impact of tasks made). not fully there yet, but confident the architecture will work in the long run

[Mathias Kleverud @mathiiias]

imo they all need to be combined

(and i see sims as subset of event-driven, we use it as a tool to produce sensor data)

just relying on one or the other never produces good outcomes, good product management is knowing when to listen to feedback, when it should shape strategy and when it should be ignored

[Mathias Kleverud @mathiiias]

judgment is about to scale

[Paweł Huryn @PawelHuryn]

AI didn't make ideas the bottleneck.

It made output free. Judgment stayed scarce.

Execution used to be the filter. AI removed it. Now anyone can produce code, copy, designs, whole products at near-zero cost. My agents produce more in a day than I used to in a month. Output exploded.

The thing that separates good output from noise didn't scale with it. Taste is knowing what's worth making, what to cut, and when it's done. It never got cheaper.

The crowd rushed to the side that got easy. The moat was on the side that didn't.

[Mathias Kleverud @mathiiias]

agent looping is agent leadership

you can micromanage one agent/human 1:1

or

you can provide the goals, the guardrails, the principles and the structure where high-level goals and intent is enough for the agents to figure out the right set of actions

[Matt Van Horn @mvanhorn]

https://t.co/DM0CAuyprS

[Mathias Kleverud @mathiiias]

we've built @momentalos on "loops" - but the big loop, where the goal is a business goal and verification is measuring what tasks actually have an impact on metrics, and then doubles down on what works

it forces prioritization of how tokens are spent (it's simply the prod mgmt loop), and we also have a budget cap on each goal ("this goal is worth x monthly")

[Mathias Kleverud @mathiiias]

figuring out what's between us and reaching our goal.

> suggestions to review and edit
> grounded in our company context
> smooth ui

[Mathias Kleverud @mathiiias]

cloud agents + local agents + humans, just drag and drop to assign the task (or outcome!)

[Momental @momentalos]

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.

[Mathias Kleverud @mathiiias]

🎯

[Bloomberg TV @BloombergTV]

"We already have artificial general intelligence," says Databricks CEO Ali Ghodsi.

"We don't need AI to get smarter, it is just lacking context" https://t.co/92hsCH9nON

[Mathias Kleverud @mathiiias]

shared obsidian meets miro/trello for agents and humans might sound messy but its pretty neat

[Mathias Kleverud @mathiiias]

please checkout https://t.co/jG7vJdiEhW!

it's shared obsidian meets miro/trello for agents and humans

you set a goal and get help breaking it down into what needs to be done based on your memory and context

drag and drop our cloud agents, your mcp connected agents (claude code, hermes etc) or humans (or your human teammates' agents!) to put them to work

steer work and get updates on whatsapp, pull in data from ga and ship directly to github

[Mathias Kleverud @mathiiias]

please checkout https://t.co/jG7vJdiEhW!

it's shared obsidian meets miro/trello for agents and humans

you set a goal and get help breaking it down into what needs to be done based on your memory and context

drag and drop our cloud agents, your mcp connected agents (claude code, hermes etc) or humans (or your human teammates' agents!) to put them to work

steer work and get updates on whatsapp, pull in data from ga and ship directly to github

[Mathias Kleverud @mathiiias]

coming soon? live at https://t.co/6y6EgL3IZg now

[Perplexity @perplexity_ai]

Today we're announcing that hybrid agentic inference is coming to Perplexity Computer.

Computer can split tasks between a local model running on your machine and frontier models in the cloud. This keeps private data on your device and maximizes token efficiency.

Coming soon.

[Mathias Kleverud @mathiiias]

check out https://t.co/jG7vJdiEhW! we built it based on how the best product teams work, for agent-first teams

it comes with
> memory - shared between you and agents and any human you later bring on
> planning - you set an objective, momental helps you break it down into tasks, and you can view them in task boards
> agents - drag and drop our managed agents to tasks, or assign your claude code - they'll pick up tasks with full context in vs code / terminal
> connections - connect ga, posthog, mail and agents will use those too

stay in full control or set it to full autonomous and they'll keep working on the next best thing

also, if you're a one man show, you can steer your whole workspace from whatsapp which is pretty nice

[Momental @momentalos]

You started a company to build something great.
Not to do everything yourself.

Starting today, Momental makes any company self-driving.

Connect your GitHub and analytics, set a goal, and Momental gets you there.

For real businesses. With real products, real customers, and real data that actually matters.

[Momental @momentalos]

this is a visual representation of we work with agent memory

[Mathias Kleverud @mathiiias]

@nikunj Sent you a DM, re autonomous teams.

[Mathias Kleverud @mathiiias]

We recently created an online bookstore that runs itself and blogs about the decisions and actions it takes.

Below I’m sharing the behind the scenes - how we gave it an objective and team of agents running in the background.

[Momental @momentalos]

Meet Allo - a little pixel guy is running his own online bookstore.

His mission: "Make the world read more." His objective: sell as many books as possible.

He has six bookshelves, can only change his selection every 24 hours, and blogs about every decision he makes. No script or predefined workflow. We genuinely don't know how he'll solve it.

This weekend: fixed some bugs, made his first sale, and figured out his real problem - nobody knows his store exists.

He's working on it.

→ https://t.co/cWm7MrVjt3

[charlotte @lottite]

We didn’t want to build a memory layer, we just wanted a really intelligent team of agents.

Turns out there’s no shortcut.

[charlotte @lottite]

I’ve been saying that we with @momentalos give agents the shared context and tools they need to actually work

but I realize I maybe should say

Momental is a frontier applied AI lab building a post-agentic organizational intelligence substrate that leverages multi-modal epistemological context graphs to synthesize structured epistemic primitives into a dynamic, conflict-aware, semantically-resolved context mesh - orchestrating autonomous agent swarms through a principle-grounded agent harness to deliver operational superintelligence at civilizational scale - purpose-built for the autonomous enterprise of tomorrow.