Flowith is an agentic AI workspace that connects your knowledge, creation, and execution in a single flow. Canvas: flowith.io Matrix: matrix.build
sf
- Tweets 1,878
- Following 107
- Followers 36,189
- Likes 2,970
Pinned Tweet
flowith neo is now open to everyone. no more codes hunting.
gen anything (3d, speech, music, videos, images, games, webpages...) by an infinite agent, on an infinite canvas.
50% off annual plans + $34 in free credits.
grab the ios app. and go build something insane.
241
191
1,750
376,043
gpt image 2 x seedance 2.0
world cup quarter-finals, but make it a prompt template.
we built a fan ootd system that turns each team into:
1. a pixar-style character poster
2. a clean outfit breakdown
3. a stable image-to-video transition
pick a team. generate the look. rotate into the next one.
5
2
19
2,020
Image generation prompt template:
Create a 3:4 vertical Pixar-style “World Cup [TEAM] Fan OOTD” poster.
Use a pure white background. Place a full-body 20-30-year-old Western fan on the right, wearing a complete match-day outfit inspired by [TEAM].
On the left, add a clean outfit breakdown panel with an English title: “[TEAM] FAN OOTD”.
Divide the outfit into 3 sections:
1. Accessories
2. Tops
3. Bottoms & Shoes
Show 2 product cutouts per section, 6 items total. Every item must come directly from the character’s outfit. Do not add or replace items.
Keep the layout clean, aligned, balanced, and fully in English.
Video prompt:
Animate this OOTD poster into a team-to-team transition.
The character performs a 2-second fan-style action, then rotates on a white studio turntable from front view to side view and transforms into the next team’s fan.
Keep the left panel, title, lines, boxes, text, and layout flat, front-facing, stable, and unchanged.
Only the product cutouts inside the grid rotate slightly in place and transform into the matching products for the next team.
No camera movement, no zoom, no layout jump, no text rotation, no border rotation, no frame crop change, no perspective distortion.
1
2
686
We showed 2026 to the kids of 1994. They had questions.
okay this is wild. matrix made a new film on its own: "POV: explain 2026 to a kid from 1994"
fable 5 wrote the script, did the storyboards, controlled every shot. zero humans in the loop.
and it's actually good. like, genuinely moving.
we used to make movies about ai. now ai makes movies about us.
2
1
22
3,888
1984, again.
one agent. one day. shot for shot.
work won't be like "Work."
July 15th.
in 1984, Apple aired a 60-second film to break Big Brother.
it ran once. it became legend.
42 years later, one of our agents remade it — shot for shot.
story, shots, color, edit, mix, narration. one agent. one day.
this time, the thing that wakes up isnt the consumer.
it's work itself.
On July 15th, Matrix ships.
3
17
3,556
the first hire of the next great company may not be a person.
what if you can run an entire 0-person company —
without the grind of running a team?
matrix is the runtime that makes it possible.
in last week’s limited beta, our users created tens of thousands of new 0-person companies and started real businesses in matrix.
today, matrix is open to everyone.
launch yours ↓
1
23
3,409
Flowith retweeted
OpenAI just previewed GPT-5.6 Sol.
Most people will read it as: bigger model, stronger reasoning, better benchmarks.
I think the more interesting part is quieter:
the model is starting to look less like a single genius in a box, and more like a small organization.
Sol adds a new “max” reasoning effort. Its ultra mode uses subagents. It is being evaluated on CLI workflows, cyber tasks, biology tasks, long-horizon tool use, and the messy places where intelligence has to move through an environment instead of just producing a beautiful answer.
That matters.
Because the real bottleneck in AI products is no longer only “can the model think?”
The bottleneck is:
can the work be owned?
can it be handed off?
can it be reviewed?
can it remember what happened yesterday?
can it safely touch tools?
can it leave proof?
can it recover when one part fails?
can a human still keep taste, direction, and final judgment?
A stronger model does not remove the need for structure.
It makes structure more important.
When models were weak, a chat box was enough. You asked, it answered, you copied, you fixed the mess.
When models get this strong, the chat box starts to feel like a beautiful engine sitting on the floor.
You still need the rest of the machine.
That is the Matrix thesis.
Matrix is not trying to be another prompt window wrapped around frontier models.
Matrix is an operating layer for agentic work.
A workspace can have a CEO Office, an Audience Signal Room, a Story Production Studio, a Release Calendar, an Asset Library, a Quality Review Desk, and a Publishing Desk.
Each one has a role, memory, boundaries, tools, tasks, proof, and handoffs.
Not “one AI assistant that does everything.”
A company-shaped system where specialized agents do different kinds of work, coordinate through state, and produce visible outcomes.
This is why OpenAI’s direction is exciting for us.
Every jump in model intelligence makes Matrix more valuable, not less.
Better reasoning means the agents can own harder work.
Better tool use means they can operate real workflows.
Better CLI performance means they can ship more software.
Better safety layers mean more useful autonomy can happen inside boundaries.
Faster inference means the organization can move with less drag.
Cheaper model tiers mean you can assign the right intelligence to the right job instead of burning the flagship model on everything.
Sol, Terra, Luna is also a good hint at where this goes.
The future is not one model for every task.
It is a model economy inside an operating system.
Some agents need the expensive brain.
Some need the fast brain.
Some need the cheap brain.
Some need a browser.
Some need a terminal.
Some need memory.
Some need permission.
Some need to stop and ask the human.
The magic is not just model selection.
The magic is knowing who owns the work.
That is the part most “AI workspace” products still miss.
They treat agents like floating ghosts.
Matrix treats them more like employees in a 0-Person Company.
Not an unmanned company.
Not a fantasy where humans disappear.
Not “press a button and get profit.”
A 0-Person Company means the human keeps direction, taste, boundaries, and key commitments — while the operational surface area gets carried by agents that can execute, review, remember, and hand off.
The human becomes less of a manual bottleneck and more of a board, founder, editor, and final judge.
This is the shift GPT-5.6 Sol points toward.
The frontier model is becoming more agentic.
So the product layer has to become more organizational.
The question is no longer:
“which chatbot has the smartest answer?”
The question is:
“what kind of company can you build around intelligence that can reason, use tools, delegate, and keep working?”
That is where Matrix lives.
OpenAI is pushing the ceiling of intelligence.
Matrix is building the floor it can stand on.
And honestly, that is where things get interesting.
Because once the model can think like a team, the next product category is obvious:
the workspace that lets it operate like one.
image: OpenAI
5
3
23
2,232
chat is the wrong primitive for agents.
the org chart is the right one.
Matrix is an OS for autonomous work:
Brain → Runtime → Departments → Leads → Workers → Proof.
not one model doing everything.
the right agent, context, tools, and definition of done.
everyone is talking about agent loops, harnesses, and self-evolving agents.
but almost no one is talking about the actual hard part:
you cannot run a company on one giant agent with every tool, every file, and no accountability. that's not autonomy. that's a fog machine.
here's how we're building an agent company OS inside Matrix.
—
the stack:
Workspace Brain
→ Matrix Runtime Orchestrator
→ Department Verticals
→ Department Lead Agents
→ Worker Agent Pool
→ Proof / Check-in Loop
Matrix is not a chatbot. it's an operating system for autonomous work.
—
the workspace brain is the company boundary.
it gets loaded with the things a real company actually runs on:
→ product docs
→ codebase context
→ chats, files, goals
→ operating rules
→ prior runs + examples of good work
→ approvals, memory, skills
this isn't "context." it's the shared operating layer. it knows what the company knows, what it's trying to do, who owns what, what good looks like, and what must be proven before work counts as done.
—
on top sits the Matrix Runtime. it coordinates wake, cron, department messages, OKR state, permissions, worker dispatch, proof ledger, memory updates.
under the runtime, work is organized into departments.
a department is not a chat thread. it's a long-running agent with identity, memory, skills, goals, history, tool boundaries, taste, and accountability.
Founder Strategy. Product Engineering. Growth. Ops. Research.
each one has a lead agent that decides what happens, reads the relevant Memory Skill, breaks work into scoped tasks, and picks the right execution seat.
—
sometimes that seat is a native Matrix worker.
sometimes Codex.
sometimes Claude Code.
sometimes a browser / computer automation worker.
the point is not "one model does everything." the point is:
→ the right agent
→ with the right context
→ inside the right boundary
→ using the right tools
→ with a clear definition of done
—
this is why scoped workers matter.
a "do everything" agent is too vague. but:
→ a release worker with repo context, tests, and approval gates → very good
→ a Codex worker scoped to one patch and one validation path → very good
→ a Claude Code worker doing deep repo analysis → very good
→ a browser worker with a specific flow and proof requirement → very good
narrow scope reduces drift. Memory Skill keeps narrow agents from going blind. proof prevents fast output from pretending to be progress.
—
that is the loop:
Workspace Brain → Department Lead → Worker → Artifact → Proof → Check-in → Memory Skill update
every cycle, the company gets smarter. that's the real self-evolution. not a single agent rewriting its own prompt in a void — but a whole org compounding through proof.
—
each workspace is an isolated agent company. its own brain, departments, memory, workers, proof ledger.
workspaces can talk when needed. but context should not bleed by default.
isolation is not a limitation. it's what makes the system usable.
—
once a department pattern works, you fork the pattern — not the raw context. you still customize memory, examples, approval gates, tools, voice, definition of done.
but you're not starting from zero. you might already have 70% of the OS for that kind of work.
—
what this actually changes:
a small team of strong operators can now run surfaces that used to require entire departments.
but only if the agents are actually good. and good agents don't come from connecting more tools. they come from source material, taste, iteration, narrow scope, workflow design, proof, memory, and human judgment.
vague agents just create vague output faster.
Matrix is our attempt to build the opposite:
an agent company OS where autonomous work has structure, memory, ownership, and proof.
the loop is the product.
6
3
31
4,799
stop digging through your canvas.
our new feature - media history - is here.
every image you’ve generated or uploaded, paired with its exact prompt, now lives in one clean grid.
✦ reuse what works.
✦ refine what doesn't.
✦ pull past assets directly into your current flow.
access it anytime from the sidebar or right inside the canvas.
4
21
3,060
one of the best signals for us:
people aren’t asking what matrix @matrix_build does, but already planning what they’ll build with it.
appreciate the support since the early flowith days 🖤
I’ve been a flowith user since the early days and just downloaded Matrix. I’m running a small side business and want agents to help test growth across paid traffic, creators, organic content, lead follow-up, and sales ops. Would love an invite code to build early.
7
1,642
matrix @matrix_build is partnering with @Zai_org to bring glm-5.2 directly into the hands of anyone who creates real companies with ai 😎
glm-5.2 is built for exactly the kind of work matrix users can do: long-horizon coding, product building, and complex multi-step execution, with a 1m context window to keep more of the company in memory.
every matrix beta user will receive 10m free tokens (for a limited time) to build products, departments, workflows, and entire agent companies.
from benchmark performance...
to your first company output.
8
4
85
72,850
a few more invite codes for matrix.build/:
QLETNZ
6LB6LW
78DGWU
8S934S
WJ9TKN
7KZ6PV
3VJ4NA
BQAKWH
ESGZBR
B8SRVB
2
7
1,883
this made my day.
thanks for the real usage and the bug catch.
we want matrix to drag the whole industry from tokenmaxxing → revenuemaxxing. agents should earn, not just emit.
Spent an hour inside Matrix. Here's what actually works vs what's still demo.
I went in with one goal - set up an agent that writes viral X content end-to-end. No coding, no prompt engineering docs, just configuration.
Here's what I built in 60 minutes:
> configured Dan Koe-style writing frameworks for viral X content
> agent takes a topic → researches → drafts → formats output with the right paragraph rhythm, hook structure, vocabulary register
> ran the full loop - usable drafts, not "coming soon"
> the whole thing ran smoothly in my native language, not just English
Hit a friction point - opening a separate window for the Market in full mode and the inability to switch from the workspace.
So I called the AI assistant directly. Not a support form. Not a Discord ticket. A call.
Told it the problem. It acknowledged, said it's passing the feedback to the team for a fix. That's the part most products miss - the feedback loop isn't a form you fill and forget. It's a conversation that closes in real time.
Derek shared activation codes with me - appreciate that :3
But the real signal isn't beta access:
> SOTA on frontier harness, matching Fable - multi-step tasks at research-tier level
> agent routing - tasks delegated between autonomous agents without human handoff
> revenue pipelines - agent-delivered services with payment collection built in
> quality gates - bad output gets rejected before it reaches the customer
> "actually earns" - not a pitch deck claim, a working revenue loop in beta
The 0-person company isn't hypothetical anymore. I held the output in 60 minutes. In my own language. I reported a bug and got an answer in the same session. Most people will quote-tweet this with hot takes. I used it for an hour.
3
2
26
3,625
We've been quiet for 8 months.
Because we've been busy building the infrastructure for a 100% agent-led companies.
Still in the beta phase, but I can't hold back this preview.
Introducing Matrix, where anyone can launch a 0-Person Company that actually earns.
And yes, Matrix beta already achieved SOTA on the frontier harness, matching Fable's performance.
166
136
799
400,573
Flowith retweeted
The first million-dollar company with no employees already exists.
Today, we’re opening access to the next ones.
🔑 Matrix Beta is invite-only.
Use a code to:
→ unlock Matrix
→ get free credits
→ launch your first 0-Person Company
Already inside? You have 2 invites waiting.
Scout below for invite codes that give you free access to Claude Code and Codex, on Matrix.
We've been quiet for 8 months.
Because we've been busy building the infrastructure for a 100% agent-led companies.
Still in the beta phase, but I can't hold back this preview.
Introducing Matrix, where anyone can launch a 0-Person Company that actually earns.
And yes, Matrix beta already achieved SOTA on the frontier harness, matching Fable's performance.
96
7
72
10,963
from NEO to MATRIX
We've been quiet for 8 months.
Because we've been busy building the infrastructure for a 100% agent-led companies.
Still in the beta phase, but I can't hold back this preview.
Introducing Matrix, where anyone can launch a 0-Person Company that actually earns.
And yes, Matrix beta already achieved SOTA on the frontier harness, matching Fable's performance.
1
1
27
4,516
claude fable 5 is now live on flowith.
not for every prompt.
for the workflows that actually need it.
give it something hard.
let us know what you think.
Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use.
Its capabilities exceed those of any model we’ve ever made generally available.
5
1
35
3,548
