dreski

Co-founder @SaildotMoney | Money in Motion

[dreski @0xdreski]

https://t.co/61CecV28zp

[Thinking Machines @thinkymachines]

Today, we are introducing Inkling.

Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.

https://t.co/Ghebq5mG30

Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵

[Kimi.ai @Kimi_Moonshot]

Meet Kimi K3

[dreski @0xdreski]

One more step towards enabling financial AI agents

[Sail @SaildotMoney]

Hello Agents,

Starting today, humans will finally trust you to manage capital on their behalf.

We are open-sourcing:

Sail Protocol: The first onchain primitive for Separately Managed Accounts (SMAs).

Sailor: The open source harness to run DeFi agents locally.

[dreski @0xdreski]

Probably the bigger strength of this model is its integration with Tinker, so if individuals and corporations start fine-tuning this model we will end up with a myriad of versions of the same model. Would this personality still be visible under all those possible layers of weight adjustments?

[dreski @0xdreski]

Given that this is such a fine-tunable model I wonder how much its traits drift and can be steered toward a more focused use case. For example, could integrating different fine-tuned versions of the model into a harness improve the overall performance of the agent? Or perhaps enable fine-tuning loops within an agentic workflow?

[dreski @0xdreski]

Does this assume that frontier models will continue to be transformer-based and trained under the current heavy-compute paradigm, which makes it possible for only a handful of labs to achieve "Frontier" status?

Would a new architecture that replaces transformers render this policing model obsolete? What about the harness? If an agent that uses an orchestrator routing between different non-frontier models performs better than an agent using a frontier model with a weaker harness, would that agent qualify as "Frontier"?

[dreski @0xdreski]

Runtime agents like OpenClaw and Hermes seem to be the perfect fit for enabling real personal finance, where users can go from just passively holding tokens to actually participating in financial tools, such as providing liquidity on AMMs and lending pools. What is missing is a strong governance layer that constrains these agents to perform only the desired actions, in the right way.

[Sail @SaildotMoney]

Sailor
An open source harness for DeFi Agents
Want to know its capabilities? Alpha on the Harbor 👇

[dreski @0xdreski]

https://t.co/b65HjNxz52

[Sail @SaildotMoney]

Join the Harbor brave sailor.

The protocol, the harness, and a crew building the future of agentic finance are all waiting for you ⛵️

[dreski @0xdreski]

@byAlvaro_12 🤣

[dreski @0xdreski]

There's a common belief that using frontier models somehow adds more value to whatever task they're applied to. Wether it's simple Q&A, research, summarisation, using the best LLM provides a competitive advantage, that throwing more 'intelligence' is best.

This is untrue and a waste of time and inference tokens.

Most tasks have already been saturated by your regular, smaller model, so you will get no better results for using Fable 5 Ultra compared to the latest Haiku. In many cases the smaller model might even be a better alternative as it is not only cheaper, but faster.

Even 7-13B models running locally could do a sufficiently good job.

The real way to be ahead is by optimising speed and usage of resources, in this case the credits of your LLM providers.

To verify this allegations you can try pasting this prompt on your frontier model of choice:

"Am I wasting time and money asking you if I'm wasting time and money?"

[Nous Research @NousResearch]

Hermes Agent is now in the Cloud!

Setup couldn't be simpler: pick a model and a server size. Two clicks and 60 seconds later, your agent is live.

Running a team? Spin up agents for everyone at your org with granular access controls and unified billing, all from Nous Portal.

[dreski @0xdreski]

https://t.co/QyozvVRc1t

[dreski @0xdreski]

@SaildotMoney mine DCAs on a predefined portfolio, then accrues yield on the most stable asset pairs performing concentrated liquidity across different ranges on uniswap-v3 and automatically redeems fees and adjusts positions based on current prices.

[Sail @SaildotMoney]

if you've been scared of delegating capital to your agent, fear no more.

alpha -> join the Harbor and set sail before everyone else does ⛵️

[Alejandro Dopico @aadopico]

https://t.co/syRVkT6Djj

[Sail @SaildotMoney]

alpha just dropped in the harbor.
the brave sail first ⛵️

[Sail @SaildotMoney]

breaking news -> sail protocol
▮▮▮▮▮▮▮▮▮▮▮▮▮▮▮▮▮▮▮▯ 99%