Sebastian Maniak

Follow me and let’s build some AI agents

[Sebastian Maniak @SebbyCorp]

This is to show many harness support. But yes different workstrees would be a great solution

[Sebastian Maniak @SebbyCorp]

agentdesktop 0.5.0 is out.. let's see it in action! In this demo I have agentdesktop running with:
- @Anthropic Claude Code, Claude Desktop
- @OpenAI Codex/GPT App - SpaceX Super Intelligence Grok Build
- https://t.co/48kQ8bXkyR Harness
- @opencode v2 Harness

Using @NVIDIA #Qwen3.8 and of course the frontier models from OpenAI and Anthropic

Link: https://t.co/BS3IAYgOjG https://t.co/4JFvHMyIHK

[Sebastian Maniak @SebbyCorp]

@johnwrightai I thinks it’s great.

[Sebastian Maniak @SebbyCorp]

Day1 Building AI Homelab with my DGX SPARK + agentgateway + pi harness

https://t.co/5TDbzMt5Hh
https://t.co/WzDUZsdDIR

[Sebastian Maniak @SebbyCorp]

Funny my AI Agent accidentally caused an outage in my home network while trying to disable Netflix and YouTube access. lol. Took my whole internet down.

lol.

[Sebastian Maniak @SebbyCorp]

Let’s goooooo

[Floro S. @sflorimm]

Revenue: $16.3B
Net loss: $80.7B
IPO valuation: $2T

[OpenAI @OpenAI]

Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.

On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.

While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.

After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.

https://t.co/EmEZsFRMcF

[Sebastian Maniak @SebbyCorp]

@pjausovec 100% it already can

[Ashish Rajan 🤴🏾 @hashishrajan]

Just another week in cybersecurity. 🚬💀

[Sebastian Maniak @SebbyCorp]

Slowly moving torwards local models, using my qwen3.8 to setup my @opencode harness

[Sebastian Maniak @SebbyCorp]

Day 2 @nvidia #dgx ai home lab with tokenomics with @agentgateway + @opencode harness

[Sebastian Maniak @SebbyCorp]

What effort do you use? Medium seems to be perfect for me with what i do..

[Sebastian Maniak @SebbyCorp]

If you dream it, you can ask AI to build. @TheNJDevOpsGuy :D

[Sebastian Maniak @SebbyCorp]

@FE80CC1E i miss u man.. thanks for sharing

[Sebastian Maniak @SebbyCorp]

You know it’s been almost 6 months since mythos was announced! So that means all the open source models have probably caught up.

[Sebastian Maniak @SebbyCorp]

Enterprises by the end of 2027 will be rolling out local models on their employee laptops. Curious what harness they will use? Is there a need for a “enterprise” “certified” harness wrapper around a https://t.co/WzDUZsd5Tj? Or pydev ?

[Sebastian Maniak @SebbyCorp]

wtf

[Sebastian Maniak @SebbyCorp]

@UK_Daniel_Card @grok hack this website. Please