Viewing profile — alexbuiko
alexbuiko
HN member- Joined
- Thu, Mar 05, 2026, 12:22 PM UTC
- HN karma
- 1
- Public activity
- 9 items
- HN profile
- View on Hacker News ↗
About alexbuiko
Developing SDAG (Systematic Defect Awareness & Guidance) to bridge the gap between model-level routing decisions and physical hardware reliability. Interested in MoE stability, tail latency analysis, and semiconductor degradation modeling at scale.
GitHub: https://github.com/alexbuiko-sketch Open for deep technical discussions on AI infra.
Recent public activity
-
comment
Comment #47348359
sent e-mail to you
-
comment
Comment #47339198
Focusing on 'Cost per Outcome' rather than 'Cost per Token' is a vital shift for AI reliability. At SDAG [ https://github.com/alexbuiko-sketch/SDAG-Standard ], we’ve been looking a…
-
comment
Comment #47339106
Exactly. What you describe as 'parsing work' is, at the architectural level, a high-entropy search across the attention heads. When a prompt is a 'wall of text,' the model's routin…
-
comment
Comment #47338849
Those sigma numbers are incredible—dropping variance by 24x practically confirms that you’ve managed to 'trap' the model in a low-entropy state. In production, predictability (the …
-
comment
Comment #47327028
This is a brilliant breakdown of the 'Token Mix' paradox. It aligns perfectly with what we’ve been seeing while developing SDAG. When you optimize for a structured context payload …
-
comment
Comment #47326958
A decade since AlphaGo, and we’re still just scratching the surface of model alignment and efficiency. While DeepMind proved that AI can master intuition, we are now looking at the…
-
comment
Comment #47272196
[dead]
- story
- story