Pinned Tweet
Replying to @florioshi
Fruth 0.1.9 is out. Yes, already.
But nothing flashy. Boring. I'm quite pleased.
Repo in the next post.
An AI response can be more than text. It can be a traceable chain of work.
Fruth 0.1.8 hardens that core through Hypothesis and systematic testing: preserving intent, enforcing authority, and binding completion to evidence.
Repo in the next post.
Fruth connects interpretation, execution, saved artifacts, and review—making AI-driven work more auditable, including across interruptions and continuations.
github.com/fl0ri0/fruth
I kept trying to break Fruth’s idea of “done”.
Wrong filename. A branch using another branch’s result. Conflicting evidence disappearing during recovery.
Audits and Hypothesis helped find these bugs.
The fixes are in Fruth 0.1.6.
Repo link in the next post.
Fruth 0.1.5 is out.
In Fruth, a response brings interpretation, execution and verifiable results together. This maintenance release strengthens the foundations: more reliable scheduling and stricter dependency validation.
github.com/fl0ri0/fruth
Fruth v0.1.4 is out.
SQLite now handles response indexing more efficiently, preserving complete history and evidence.
Both Apple PCC shortcuts are bundled in the source tarball, with assisted first-use setup.
github.com/fl0ri0/fruth/rele…
Fruth 0.1.3 is out!
The inference engine keeps evolving: better follow-ups, artifact continuity, fallback and research evidence tools.
Plus native PDF/OCR, Apple Foundation Models and Private Cloud Compute (PCC) support.
github.com/fl0ri0/fruth/rele…
Yours Fruthfully,
Florio
Fruth: A durable semantic state inference engine for local AI.
Model output alone does not prove execution. “Done” is a gate, not a sentence.
Fruth turns intent into explicit obligations and durable, inspectable work:
• Semantic Intent: Opens a space of possible work.
• Interpretive Inference: Derives and shapes candidate work graphs from current intent and evidence.
• Obligation Promotion: Promotes validated candidates into explicit, owed contracts.
• Late Fill: Carries pending obligations forward, binding saved outputs and evidence back into the same response.
• Evidence-Gated Closure: Checks fulfillment against runtime evidence before state freeze, keeping unresolved work visible.
Interpretive inference proposes. Runtime contracts govern.
Local-first. Runs directly on the host with compatible Ollama, MLX, and llama.cpp backends. No Docker or containers required.
🔗 fl0ri0.github.io/fruth/
#LocalAI #OpenSource
Injected multiple locals models (ollama + mlx) into @openclaw (by code) so it can handoff tasks not only to different subagents, but subagents with different models. All at the same time. Incredible! ofc it cleans them up too when none is available. :-)
next step i think is full integration of my local multi-model-manager into @openclaw. so it can not only choose models that are running, but start and stop the full army of local models at will.
FloriO retweeted
This petite tower behind me is the most powerful financial institution in the world. 🏦🦋🍄
It’s the Bank for International Settlements (BIS), the central bank for central banks.
Swiss authorities cannot enter this building without permission. The BIS is protected from some Swiss legal actions and is exempt from some taxes.
Here, representatives from central banks around the world meet regularly to share information and coordinate policies, especially during financial crises, and only those with an official invitation can enter.
One of its biggest scandals was during World War II, when it accepted gold from both the Allied and Axis powers, including gold stolen by the Germans. After the war, they continued their operations as if nothing happened.
Some say they even planned the 2008 financial crisis or that it was built low-key on purpose to avoid attention.
Does the BIS show why we need decentralized finance instead of central banks controlling everything behind closed doors? 🤔 – at Basel, Switzerland
1 year ago, $Cult DAO passed Prop 186 to deploy treasury funds into #Ethena. Today, we release the Annual Review.
Status: We have transitioned from manual execution to an automated on-chain liquidity providing operation.
Here is the infrastructure upgrade and the results from the first 14 days of live operations. 🧵👇
The Decision:
Per the proposal mandate, the trustees have reviewed the operations.
Verdict: CONTINUATION.
We remain in Phase 1: Accumulation. We are compounding 100% of yields to rebuild the capital base. Once the target is hit, we flip the switch to Distribution & Burn.
Trust Code, Not Promises.
Verify the engine yourself on the live transparency dashboard:
wearecultdao.github.io/trans…
Safety over Complexity.
#ForTheMany #CULT #RealYield #DeFi







