Saylor

just a boy enjoying his agentic swarm.
Stay databased.
I like ACID.

[Saylor @kruschdev]

people are finally realizing
https://t.co/ltAh8fzTKU

[Saylor @kruschdev]

@BiscuitsJesus

[Saylor @kruschdev]

@BiscuitsJesus Their lighting designer kills it too. I was really impressed with the lights both times I saw them this year.

[Saylor @kruschdev]

Coming from five years of field electrical and mechanical work, that manual layout is where I learned it too.

Nobody memorizes a 900-page codebook, you keep the master index handy and pull the exact section for the job in front of you.
I built my filesystem MCP the same way so the agent inspects a structural page tree first and only hydrates the specific page and pinned notes it needs.

[Saylor @kruschdev]

The only safe way I have found to give an agent inbox access is breaking that capability chain in code rather than system prompts. My Gmail MCP searches threads and stages drafts (sending defaults to draft-only), while a frozen policy floor blocks credential calls and cuts off arbitrary outbound HTTP or browser navigation whenever untrusted email text enters context. Give one loop raw inbox read plus a headless browser and a single poisoned email owns every account you have.

[Saylor @kruschdev]

@quantizor https://t.co/5U3LFm82gm

[Saylor @kruschdev]

@ZacariahHeim @quantizor hi I created this
https://t.co/ltAh8fzTKU

[Saylor @kruschdev]

@quantizor own your memory layer.

[Saylor @kruschdev]

@quantizor @Steve99163 https://t.co/ltAh8fzTKU

[Saylor @kruschdev]

@quantizor @Steve99163 that's why my memory system has temporal decay on memories

[Saylor @kruschdev]

@chasewoolard I'm gonna be in SF this weekend for the 8090 hackathon. Anyone I should meet?

[Saylor @kruschdev]

@AmbroseBurnyr Haha I love a good YEM with an extended vocal jam to make the bar really uncomfortable

[Saylor @kruschdev]

@AIDevEngineer @icanvardar hi, I might have something you might be interested in.

https://t.co/ltAh8fzTKU

[Saylor @kruschdev]

@icanvardar https://t.co/ltAh8fzTKU

[Saylor @kruschdev]

Driving up from LA to Redwood City this Friday for EVOLVE at @8090_Factory.

20 hours overnight to build an autonomous Software Factory with AI agents. Generating code is getting fast, so I'm focused on the verification, call-graph, and deterministic governance layer that makes AI-generated changes safe to merge in production.

Looking forward to building alongside the 8090 team and everyone in the room.

[Saylor @kruschdev]

@graphonfounder @clarkzhango @deepakmishraVC @GraphonAI Just sent you an email, Arbaaz! Looking forward to connecting while I'm up in SF.

[Saylor @kruschdev]

@clarkzhango @graphonfounder @deepakmishraVC @GraphonAI https://t.co/ltAh8fzTKU

[Saylor @kruschdev]

Congrats on the launch @clarkzhango! I'll be up in SF this week for the 8090 hackathon and share the exact same obsession with persistent agent memory. Tried to DM you but your inbox is closed. I'd love to swing by the office and talk collaboration while I'm in town. Does Friday or Monday work better for you?

[Saylor @kruschdev]

@Wbackdown @NousResearch Both. For code, it catches it deterministically: I scrub org and region names to NeutralOrg_1, score both ASTs for security sinks, and auto-curate the clean blind code. For prose, it verifies spin quotes character-for-character and surfaces silent omissions for judgment.

[Saylor @kruschdev]

Most AI alignment filters flatten perspective bias on prose and subtly shift code-audit severity depending on who appears to own the repository.

For the @NousResearch Hermes Agent Plugin Hackathon, I built Hermes Parallax v0.1.0: a zero-dependency epistemic triangulation and counterfactual security engine for Hermes Agent (/parallax).

Inside Hermes, the Fact-vs-Frame lens pits models head-to-head across six symmetrical US and global framing poles (including state_aligned, western_policy, us_progressive_institutional, us_populist_conservative, corporate_pr, and safety_boilerplate) with full English and Mandarin tokenization, checking every flagged spin quote verbatim against the source text.

For code reviews, the Blind Code Taste-Test scrubs over 40 organization, author, and comment identity markers to neutral placeholders, runs a blind pass on the identical AST, and measures exact severity-delta shifts. Across 168 runs on 7 models, Nous Arete-27B swept stock Qwen 6 to 0 on statutory framing and scored 100 across all 18 security counterfactual passes.

Every high-contrast trajectory flows into Dataset Forge, which exports clean DPO preference pairs while quarantining unsafe code fences, prompt-injection payloads, and symlink escapes. One install registers 5 native agent tools, 3 auxiliary model slots, /parallax slash commands, a React dashboard, and a standalone CLI and Stdio MCP bridge, verified across 10/10 pytest suites and 15/15 Hermes validator checks.

Repo: https://t.co/sFlieeGwcm

[Saylor @kruschdev]

I built Hermes Parallax for the @NousResearch Hermes Agent Hackathon. It runs symmetrical 6-pole framing audits, blind AST code security counterfactuals, and poison-quarantined DPO curation inside Hermes (/parallax).

https://t.co/BJE6RjC9jE