返回首页
最新
We’ve validated the Sigma Runtime architecture (v0.4.12) on Google Gemini-3 Flash, confirming that long-horizon identity control and stability can be achieved without retraining or fine-tuning the model.<p>The system maintains two distinct personas (“Fujiwara”, a stoic Edo-period ronin, and “James”, a formal British analyst) across 220 dialogue turns in stable equilibrium.
This shows that cognitive coherence and tone consistency can be controlled at runtime rather than in model weights.<p>Unlike LangChain or RAG frameworks that orchestrate prompts, Sigma Runtime treats the model as a dynamic field with measurable drift and equilibrium parameters.
It applies real-time feedback — injecting entropy or coherence corrections when needed — to maintain identity and prevent both drift and crystallization.
The effect is similar to RLHF-style fine-tuning, but done externally and vendor-agnostic.<p>This decouples application logic from any specific LLM provider.
The same runtime behavior has been validated on GPT-5.2 and Gemini-3, with Claude tests planned next.<p>We use narrative identities like “Fujiwara” or “James” because their linguistic styles make stability easy to verify by eye.
If the runtime can hold these for 100+ turns, it can maintain any structured identity or agent tone.<p>Runtime versions ≥ v0.4 are proprietary,
but the architecture is open under the Sigma Runtime Standard (SRS):
<a href="https://github.com/sigmastratum/documentation/tree/main/srs" rel="nofollow">https://github.com/sigmastratum/documentation/tree/main/srs</a><p>A reproducible early version (SR-EI-037) is available here:
<a href="https://github.com/sigmastratum/documentation/tree/bf473712ada5a9204a65434e46860b03d5fbf8fe/sigma-runtime/SR-EI-037/code" rel="nofollow">https://github.com/sigmastratum/documentation/tree/bf473712a...</a><p>Regulated under DOI: 10.5281/zenodo.18085782 —
non-commercial implementations are fully open.<p>HN discussion focus:
– Runtime-level vs weight-level control
– Model-agnostic identity stability
– Feedback-based anti-crystallization
– Can cognitive coherence be standardized?
Hi HN,<p>I ported the teaching OS xv6-riscv to HiFive Unmatched and got it running on real hardware, including passing usertests.<p>I've been self-studying OS internals using the MIT 6.1810 materials. After finishing most of the labs, I was eager to see what it's like to run the OS on bare metal, rather than QEMU.<p>The Unmatched may not have the latest RISC-V features, but it's well-documented, and the Rev B release has made it more affordable, which makes it a good learning platform.<p>The porting process involved several interesting challenges:<p>- Hardware Quirks: Handling things like enabling A/D bits in PTEs (the hardware doesn't set them automatically, causing page faults), proper handling of interrupts, and instruction cache synchronization.<p>- Boot Flow: xv6 expects M-mode on startup, but standard RISC-V boot flows (typically via OpenSBI) jump to S-mode. To bridge this gap, I created a minimal U-Boot FIT image that contains only the xv6 kernel. This way, U-Boot SPL handles the complex CPU/DDR initialization, then hands control to xv6 in M-mode (skipping OpenSBI).<p>- Drivers: Ported an SPI SD card driver, replacing the virtio disk driver.<p>I wrote up implementation notes here: <a href="https://github.com/eyengin/xv6-riscv-unmatched/blob/unmatched/doc/NOTES.md" rel="nofollow">https://github.com/eyengin/xv6-riscv-unmatched/blob/unmatche...</a><p>Hopefully, this is useful for others who are learning OS internals and want to try running their code on real RISC-V hardware.
I built a simple tool to monitor webhooks after getting burned by silent Stripe webhook failures.<p><pre><code> How it works:
1. Create a proxy URL in HookWatch
2. Point your webhook provider (Stripe, Shopify, etc.) to it
3. We forward to your server and log everything
If something fails, you can see the full request and replay it with one click.
https://hookwatch-production.up.railway.app
Built with FastAPI + SQLite. Happy to answer questions about the architecture.</code></pre>
I manage several HubSpot instances, and getting data from PDF contracts into the CRM is a massive bottleneck.<p>Standard OCR + Regex was too brittle. So I built Scanny AI.<p>It listens for Drive webhooks, uses a vision model to extract keys (like "Total Value") regardless of layout, enforces a strict JSON schema, and patches the HubSpot API.<p>It handles about 5k pages/hour.<p>Docs and API keys: scanny-ai.com