Mani from DeployClaw.com

Building
Your Server, Your AI, Your Data, Your Privacy, Your API, Your Agent

[Mani from DeployClaw.com @manishipsfast]

Is LLM a simulation or real?

[Mani from DeployClaw.com @manishipsfast]

@OsaurusAI I see will upgrade

[Mani from DeployClaw.com @manishipsfast]

@OsaurusAI no

[Mani from DeployClaw.com @manishipsfast]

@Alibaba_AMAP Thank you for your response. I'm going to try it out on my MacBook and get back to you.

[Mani from DeployClaw.com @manishipsfast]

With auto-translation

On Twitter

It is allowing Chinese, Japanese, or any other non-English speakers to write in their language and promote themselves.

This is huge.

Recently, I am reading a lot of Chinese, Japanese, or even Korean tweets that are being auto-translated. This is true globalization.

Everyone will feel more connected. People will start buying from anywhere in the world.

I'm seeing SaaS products being launched by Chinese. I'm seeing new LLMs being created by Chinese. All these are happening in X platform.

X-platform has opened a huge market to the non-English speakers.

[Mani from DeployClaw.com @manishipsfast]

@DagmawiBabi Does it work with only public repos or private repos as well?

[Mani from DeployClaw.com @manishipsfast]

@ikbenbasha Where can I see the demo?

[Mani from DeployClaw.com @manishipsfast]

All my life I have been a user of these LLM models. Now I'm trying to make my own LLM in my laptop.

Let's see how it goes.

[Mani from DeployClaw.com @manishipsfast]

@chriscyph @openclaw @claudeai So, SaaS to skill is the future?

[Mani from DeployClaw.com @manishipsfast]

Lee Chaton Fat

Fable killer?

[Mani from DeployClaw.com @manishipsfast]

@skirano Only design together or code together as well.

[Mani from DeployClaw.com @manishipsfast]

I think

Fusion is the future.

Mixing models and mixing harness may give you better results compared to one standalone model.

Open-router did that. I read a tweet about it. They improved the results by some X%, I don't remember exactly.

Imagine mixing hundreds of models. Let's say you give a prompt to Claude Code or Codex, and it spawns X number of sub-agents. Each sub-agent is a local model or the one through the open code. I think this is an interesting idea to try. Probably I'll try it. Let's see.

[Mani from DeployClaw.com @manishipsfast]

@cohere Any benchmarks?

[Mani from DeployClaw.com @manishipsfast]

@OpenRouter Which model is the judge model? Is it Claude?

[Patrick Jiang @patpcj]

Introducing Harness-1, a 20B search agent trained with a state-externalizing harness.

> frontier-level long-horizon search, rivaling Opus-4.6 and outperforming GPT-5.4

> Context-1-level cost and latency

> externalizes candidates, evidence, verification, and search history

> open-source

[Mani from DeployClaw.com @manishipsfast]

I have tried instantly before. It is a little expensive, but it's OK

I think the main problem is not warming up accounts or sending quality or something, because we have services like Instantly or SmartLead they take care of warming up accounts & deliverability.

I think the biggest problem here is the lead generation. Generating a consistent number of leads every day for the entire year without paying much is the big deal here.

Whom do you target if you want to sell your AI services?

[Mani from DeployClaw.com @manishipsfast]

@alex_here_now How many emails did you send?

I think we should send minimum 10,000 emails per month to see noticeable results

[Mani from DeployClaw.com @manishipsfast]

@manishipsfast playing around with it for my agent business that seems moderately similar to your thing, not having success. i think i would be doing better if i used an inbox warming service + scaled up massively, but doesn’t seem terribly worth it

[Mani from DeployClaw.com @manishipsfast]

Any SaaS founders here who used cold email for SaaS sales?