←
⚡🛡️ Evan Pappas6,909 posts
⚡🛡️ Evan Pappas avatar

⚡🛡️ Evan Pappas

@Hevalon

🛡️ Ex Technologia Libertas - Έλευθερία διὰ τῆς τέχνης - (Dec/Acc)

◷ Joined September 2009▧ 775 Photos and videos
Sam Dare avatar
⟳ ⚡🛡️ Evan Pappas retweeted
Sam Dare@DistStateAndMe·Sep 14

We are cooking some pretty interesting products around post training and inference @tplr_ai . Would love to talk to engineers and researchers to understand where you bottlenecks are . The rewards are my eternal gratitude and some credits. DM if you're up for it.

templar avatar
⟳ ⚡🛡️ Evan Pappas retweeted
templar@tplr_ai·Sep 3

Crucible, Templar's pre-training platform, has completed its first production end-to-end training runs. The latest trained an 8B model on 50.53B tokens across 48 distributed A100s, at an estimated $0.1202 per million tokens of GPU rental. The run reached 48.3% effective MFU. At AWS p4de Capacity Blocks pricing, a 48-A100 cluster operating at the literature-derived 65% compute ceiling comes to an estimated $0.1686 per million tokens. Crucible's measured $0.1202 was about 29% lower after its low-bandwidth overhead. The comparison excludes R2 storage and operations. The full writeup shows the method and accounting. Blog: tplr.ai/publications/blog/in…

0:41
Peter Robinson avatar
⟳ ⚡🛡️ Evan Pappas retweeted
Peter Robinson@drinkcoffee2010·Aug 11

Introduction to LLM Design & Covenant 72B: This week at the Ethereum & AI Engineering Group meetup I provided a 15 minute introduction to LLM design and then dived into the design and training of Covenant 72B; the largest model trained over the Internet. youtube.com/watch?v=TlVQbiTd…

⚡🛡️ Evan Pappas avatar
⚡🛡️ Evan Pappas@Hevalon·Aug 3

That shape is all over this year's agent work. Self-Harness ran one improvement across three models and got three different sets of harness changes, because each model had different weaknesses to fix. Its abstract calls effective harness design "inherently model-specific."

Post attachment
⚡🛡️ Evan Pappas avatar
⚡🛡️ Evan Pappas@Hevalon·Aug 3

You cannot write this off as a naive scheduler either. Continuum is built for multi-turn agentic serving, and its KV hit rate still fell from above 90% to about 60% once batches grew enough to create memory pressure.

Post attachment
⚡🛡️ Evan Pappas avatar
⚡🛡️ Evan Pappas@Hevalon·Aug 3

The harness knows it is running one stateful workflow. The scheduler sees independent requests. Neither is wrong.

⚡🛡️ Evan Pappas avatar
⚡🛡️ Evan Pappas@Hevalon·Aug 3

Memory gets tight. Between one model turn and the next, while a tool call is out, the inference scheduler evicts the loop's KV cache. The tool returns and the server prefills the entire growing history again.

Post attachment
Blockchain Scotland avatar
⟳ ⚡🛡️ Evan Pappas retweeted
Blockchain Scotland@blockchain_scot·Jul 28

[1/9] Edinburgh Decentralised Finance Summit 2026. Tuesday 1 September, Royal College of Physicians of Edinburgh. A full day on institutional DeFi, tokenisation, risk management, regulation and autonomous finance. Around 25 speakers. 300 attendees. luma.com/scotland-kgav

Post attachment
⚡🛡️ Evan Pappas avatar
⚡🛡️ Evan Pappas@Hevalon·Jul 8

The confidentiality fix shipped a year+ ago: a per-tenant cache_salt, assigned server-side. If the caller picks it, an attacker just picks yours. The fairness fix, reserved cache partitions, is still research. Not a stock engine flag.

Post attachment
⚡🛡️ Evan Pappas avatar
⚡🛡️ Evan Pappas@Hevalon·Jul 8

One design choice causes both: keying the cache by content and sharing it across tenants who don't trust each other. Residency becomes observable (by timing) and unprotected (from eviction) at the same time.

⚡🛡️ Evan Pappas avatar
⚡🛡️ Evan Pappas@Hevalon·Jul 8

Same root cause, second bug. That shared pool evicts by recency. A noisy neighbor's flood flushes your warm cache no matter how high your priority. Priority orders the compute queue. It doesn't touch eviction.