Joel Timana

Building things that think for themselves
Sketch → software. That's the current obsession
@fdotinc @pymutai

[Joel Timana @joevidev]

Today, for the first time,
@dianamunozlar and I are introducing @buivoai 🔥.

AI gave every role a faster way to work alone.
Business prompts. Product prompts. Engineering prompts.
Going in different directions.

The problem was never the LLMs.
It was the interface.
So we built it.

The oldest human expression now ships software.

The waitlist is open:
https://t.co/UKASApBasg

[Joel Timana @joevidev]

@shengkun_ye So good, congrats

[Joel Timana @joevidev]

@hwchase17 Yes Sir

https://t.co/jB2LGROLFB

[Joel Timana @joevidev]

Okay, here we go again
Graph or loop?

[Joel Timana @joevidev]

A remake of this was needed @juliafedorin 😂
https://t.co/Z3tdnCbKOq

[Julia Fedorin @juliafedorin]

come by Marina Green RIGHT NOW and tell us your opinion!

[Joel Timana @joevidev]

Okay, here we go again
Graph or loop?

[Joel Timana @joevidev]

@steipete For that reason, we created Hymoex (a role-based architecture for graph execution)

https://t.co/RXlQNCoQ99

[Joel Timana @joevidev]

@steipete For that reason, we created Hymoex (a role-based architecture for graph execution)

https://t.co/RXlQNCoQ99

[Joel Timana @joevidev]

@Raul_RomeroM @trainrobotsfast Sick 🔥

[Joel Timana @joevidev]

@hthieblot Half-Life of course , and the crossfire map please

[Joel Timana @joevidev]

Imagine those hands but coding 😳

[1X @1x_tech]

NEO’s Hands
An API to the Physical World

[Joel Timana @joevidev]

The best way to stress test your product is to have it used at a YC 24h Hackathon 👀

[Joel Timana @joevidev]

Paper and resources
https://t.co/N1ohyHjyu0

[Joel Timana @joevidev]

Introducing ReactLines.
Atomic Procedural Memory Units
for cognitive agents.

The problem
Agent behavior lives in giant prompts.
Instructions collide, drift, and conflict.
They can't be scoped or reused.

A formal 11-field schema
for behavioral policies
who does what, when, why,
and under what constraints.

Six scopes, from narrow to global:
TURN → SESSION → FLOW →
AGENT → ORG → GLOBAL.
Broader scopes subsume narrower ones.

A 5-stage retrieval protocol
injects only the relevant rules,
recovering 95% of them at top-3.

Built for scale:
add more rules without growing your prompt.
At 100 rules, 97% fewer tokens
than a monolithic prompt

Deployed in real-world production environments
Research designed and executed for large-scale orchestrations.

PS: Useful resources and links in the first comment

[Joel Timana @joevidev]

Introducing ReactLines.
Atomic Procedural Memory Units
for cognitive agents.

The problem
Agent behavior lives in giant prompts.
Instructions collide, drift, and conflict.
They can't be scoped or reused.

A formal 11-field schema
for behavioral policies
who does what, when, why,
and under what constraints.

Six scopes, from narrow to global:
TURN → SESSION → FLOW →
AGENT → ORG → GLOBAL.
Broader scopes subsume narrower ones.

A 5-stage retrieval protocol
injects only the relevant rules,
recovering 95% of them at top-3.

Built for scale:
add more rules without growing your prompt.
At 100 rules, 97% fewer tokens
than a monolithic prompt

Deployed in real-world production environments

Paper and resources
https://t.co/BqLgQCO9SI

[Joel Timana @joevidev]

AI chat is dead for building software...

[Joel Timana @joevidev]

Introducing HyMoEx.

Hybrid Modular Coordinated Experts
an architectural paradigm for scalable multi-agent systems
and expert coordination.

The problem:
Multi-agent systems work with 2 or 3 agents.
At 5, 10, 20, 50 or more; flat architectures break under
quadratic communication complexity.

A formal 7-role agent taxonomy
with typed message protocols.

Three deployment modalities
M1, M2, M3 with a subsumption theorem proving
M1 ⊂ M2 ⊂ M3. Migration preserves 100%
of existing agent definitions.

Mixture-of-Experts gating adapted
from neural networks to discrete agents.

96.7% expert selection accuracy.
93% token reduction from 5 agents onward.

Framework-agnostic by design.
Bidirectional adapters included.

Deployed in real-world production environments
across sectors such as developer-tools, fintech, hr, sales, deep-tech

Research designed and executed for large-scale orchestrations.

Pymut Labs
@PymutAI

Paper, Resources and links
https://t.co/06OPbAAx0a

[Joel Timana @joevidev]

Paper and resources

https://t.co/0jFwawNZgP

[Joel Timana @joevidev]

Introducing HyMoEx.

Hybrid Modular Coordinated Experts
an architectural paradigm for scalable multi-agent systems
and expert coordination.

The problem:
Multi-agent systems work with 2 or 3 agents.
At 5, 10, 20, 50 or more; flat architectures break under
quadratic communication complexity.

A formal 7-role agent taxonomy
with typed message protocols.

Three deployment modalities
M1, M2, M3 with a subsumption theorem proving
M1 ⊂ M2 ⊂ M3. Migration preserves 100%
of existing agent definitions.

Mixture-of-Experts gating adapted
from neural networks to discrete agents.

96.7% expert selection accuracy.
93% token reduction from 5 agents onward.

Framework-agnostic by design.
Bidirectional adapters included.

Deployed in real-world production environments
across sectors such as developer-tools, fintech, hr, sales, deep-tech

Research designed and executed for large-scale orchestrations.

Pymut Labs
@PymutAI

PS: Useful resources and links in the first comment

[Diana Munoz @dianamunozlar]

The rule @jowyang sets in every room:

- Unconditionally help others.

Llama Lounge 25 @Stanford, one of the best spaces in Silicon Valley to find AI founders, investors, corporates & builders all in one room.

Every series is better than the last. 🦙🐚

Were you there? Let's connect!

#LlamaLounge

[Diana Munoz @dianamunozlar]

I used to think you could build from anywhere.

Then I moved to San Francisco.

Yesterday @snowmaker nailed it at @fdotinc 💚: the best thing about SF is the people.

The smartest, most ambitious, most open minded people you’ll ever meet.

SV is the best place in the world to build!