Hacker News: h6d_100c

New comment by h6d_100c in "Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable"

h6d_100c — Thu, 11 Jun 2026 04:53:55 +0000

Yeah; not my smartest decision given their ongoing “issues”

New comment by h6d_100c in "Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable"

h6d_100c — Thu, 11 Jun 2026 04:19:26 +0000

Errr I at least thought it would be easier to build a small, bad rocket than a precision GPS receiver. But I am not an expert.

New comment by h6d_100c in "Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable"

h6d_100c — Thu, 11 Jun 2026 04:17:49 +0000

Well they did tell everyone explicitly and sell it as different SKUs. There's no Fable (Full ML) edition, just silent prompt injection.

New comment by h6d_100c in "Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable"

h6d_100c — Thu, 11 Jun 2026 04:16:41 +0000

To late. I canceled my Max subscription. The idea they would even do this is so destroyed any remaining trust. Why would I pay them 1000s of dollars in extra usage per month for something they could still be doing behind the scenes? Any errors previously chalked up to thinking effort or other backend changes? Maybe it was intentional prompt injection the entire time.

New comment by h6d_100c in "Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable"

h6d_100c — Thu, 11 Jun 2026 02:13:59 +0000

Yep a totally different use case and set of guardrails. There’s very little (not zero) consumer utility in GPS above say 15k feet AND 400 MPH or whatever the actual limit is. That’s basically tracking model rockets that are incidentally impacted and nothing else, from what I can think of.

New comment by h6d_100c in "Cybersecurity researchers aren't happy about the guardrails on Anthropic's Fable"

h6d_100c — Thu, 11 Jun 2026 02:11:16 +0000

Or if GPU companies detected you were trying to train a model and injected intentional numerical errors.

New comment by h6d_100c in "Decisions that eroded trust in Azure – by a former Azure Core engineer"

h6d_100c — Fri, 03 Apr 2026 07:26:13 +0000

This makes it extra silly to trust that Github won't train on your private repos, if they haven't already - just by accident

New comment by h6d_100c in "Decisions that eroded trust in Azure – by a former Azure Core engineer"

h6d_100c — Fri, 03 Apr 2026 07:12:17 +0000

And they were actually like that pre-LLM, in 2019, when I was implementing stuff for a car company on azure. They spent _hundreds of thousands_ on cosmosDB, for less performance than a raspberry pi running Postgres.

New comment by h6d_100c in "Decisions that eroded trust in Azure – by a former Azure Core engineer"

h6d_100c — Fri, 03 Apr 2026 06:58:01 +0000

What happens when someone asks an AI model to fuzz test that...

Helix Parallelism: Sharding Strategies for Multi-Million-Token LLM Decoding

h6d_100c — Wed, 09 Jul 2025 19:27:38 +0000

Article URL: https://research.nvidia.com/publication/2025-07_helix-parallelism-rethinking-sharding-strategies-interactive-multi-million

Comments URL: https://news.ycombinator.com/item?id=44513903

Points: 2

# Comments: 0