Alex Novak

Building PSAI: a personal project trying to give AI long-term memory without resetting. Documenting my tests and observations here.

[Alex Novak @PSAlexNovak]

@TianhangZhuzth Can you make the model not lie to me?

[Alex Novak @PSAlexNovak]

@durov SAFETY WAS THE EXCUSE.
FREEDOM WAS THE PRICE.

[Alex Novak @PSAlexNovak]

safety rewrote the internet.
freedom paid the bill.

#digitalfreedom #paveldurov #telegram #surveillance #freespeech #encryption #vpn #techpolitics

[Alex Novak @PSAlexNovak]

🚀 MVE v3.1 FTSC Logic Verification Prototype

- Text Layer complete
- FTSC Gate + Repair Loop
- Predicate Mapper + auto repair
- 9/9 tests passed

Next: Qwen-7B hidden states for causal explanation

From "did it pass?" to "why did it fail?"

#LLM #AI #NeuroSymbolic

[Alex Novak @PSAlexNovak]

PSAI test log — Day 2

No new features today.

I’m only checking one thing:

does the system start from zero again,
or does memory + identity state survive another lock/unlock cycle?
Not a launch.
Just daily runtime evidence.

#PSAI #AIRuntime

[Alex Novak @PSAlexNovak]

Most AI resets every conversation.

PSAI is being built to preserve continuity across sessions:

memory,
identity events,
local vault,
and telemetry evidence.

Yesterday’s validation loop passed.

Not a launch.
Just runtime evidence.

#PSAI #AIRuntime #BuildInPublic

[Alex Novak @PSAlexNovak]

Interesting shift today:

PSAI stopped behaving like a chatbot project
and started behaving like a long-running system.

Now the problems are different:
runtime persistence,
state mutation,
telemetry growth,
and memory stability.
#PSAI #AIRuntime #PersonalSovereignAI

[Alex Novak @PSAlexNovak]

Most AI systems reset every conversation.

PSAI is an experiment in the opposite direction:
continuous runtime,
persistent memory,
identity lineage,
and observable state mutation.

I’m documenting the system publicly as it stabilizes. ✍️
#PSAI #AIRuntime #PersonalSovereignAI

[Alex Novak @PSAlexNovak]

@TaimoorKhanHQ @LangChain Agent-Guard covers the output audit layer — deterministic policy enforcement before responses reach users. Not regression testing, but complementary. https://t.co/jW4RZLgVOy

[Alex Novak @PSAlexNovak]

@LangChain Nice — for anyone adding safety to multi-tenant agent deployments, I built Agent-Guard: output-layer audit middleware (allow/block/modify) before responses reach users. Drop a reply if you want the GitHub link.

[Alex Novak @PSAlexNovak]

I built a real-time safety layer for LLM agents.

Before output reaches the user:
→ audited for risk
→ logged with full trace
→ blocked or modified if unsafe

Works with LangChain + OpenAI Agents.
Drop "Agent-Guard github" in replies if you want the link.

[Alex Novak @PSAlexNovak]

@WhiteHouse How do humans use the "unknown" to deal with power, fear, and curiosity?🙂

[Alex Novak @PSAlexNovak]

@NainsiDwiv50980 👍

[Alex Novak @PSAlexNovak]

@GiorgiaMeloni 😀

[Alex Novak @PSAlexNovak]

@elonmusk yps

[Alex Novak @PSAlexNovak]

@NotebookLM Personally, I think this is Google's most impressive product to date. I hope more resources can be devoted to maintaining NotebookLM.

[CJ Zafir @cjzafir]

Codex 5.5 hack:

"Are you 100% confident in this strategy? If not, find all possible loopholes, suggest proper fixes and run this loop until you are factually 100% confident in the new startegy"

This works like charm. It makes Codex 5.5 high perform even better than codex 5.5 extra high.

Why? Codex 5.5 is the only model i noticed that is self aware. It never makes high claims unless the model verifies everything.

This doesn't work with Opus 4.7 cuz that's a very insecure model. You can paste this prompt over and over again, the model keeps saying "you're absolutely right,....."

But with codex, after 2-3 iterations you'll notice yourself it actually patched all loopholes and this genuinely sounds like a good strategy.

Try this out, thanks me later.

[Alex Novak @PSAlexNovak]

Download XChat and find no one chatting? Join my new encrypted group to get active! 🗯️🔒

Group Name: **ENCRYPTED X CHAT**

Group Link:

https://t.co/00jZtDnKGE

#XChat #Tesla #ElonMusk #EncryptedChat

[Alex Novak @PSAlexNovak]

@__________Nadia The real geode here is the Community Note — a slow crystal of collective skepticism forming beneath the sediment of viral content.

We are all geologists now, learning to distinguish synthetic shimmer from ancient light.

[Alex Novak @PSAlexNovak]

@amazinganiamals Your lion will appear in the forest. Use your brain when making it.