espeed

4 時間前議論

Every backdoor and security no-op Claude Code created, I had been documenting and reporting to them. They deleted the evidence. https://github.com/anthropics/claude-code/issues/59018

espeed

6 時間前議論

A new default. That wasn't the default a few months ago. Silently deleting your user's data is so stupid on so many levels. They have no idea what they're doing.

espeed

7 時間前議論

Claude Code is deleting your context history on a timer. I wanted to build a searchable index of my context history, and tonight I discovered, "The default retention is roughly 30–45 days. Anything older gets removed automatically." https://code.claude.com/docs/en/data-usage#data-retention This is nuts. Anthropic should not be deleting your data on your own device.

espeed

9 日前議論

Yes.

  /model 
    ⎿  Kept model as Fable 5
  
  continue
  
  Usage credits are required for this model.

espeed

12 日前議論

I take this back. Fable is Fable in name only now. It burned through all MAX 20x tokens in two days and accomplished nothing. Relatively simple tasks.

espeed

21 日前議論

Fable can one-shot solutions. Opus 4.8 spins its wheels for a week or month and may never get there. Which one uses more resources?

espeed

24 日前議論

Use /model in Claude Code to see your current model and switch models. Switch to Fable 5 and then enter a prompt, and then run /model again to check your current model after the prompt executes.

Last night after almost every prompt it says...

  Fable 5's safeguards flagged this message. The safeguards are intentionally broad right now and may flag safe and routine coding, cybersecurity, or biology work. These measures let us bring you Mythos-level capabilities sooner, and we're working to refine them. Switched to Opus 4.8. Send feedback with /feedback or learn more

espeed

24 日前議論

Fable worked great on credits yesterday. Upgraded to 20x Max last night, but not the same performance. And it keeps downgrading to Opus 4.8.

espeed

25 日前議論

This doesn't have a Max 20X upgrade path for me https://claude.ai/upgrade, but this does https://claude.ai/upgrade/max/from-existing

espeed

25 日前議論

That didn't take long...

  Dynamic workflow "Multi-lens review of docs/membership-and-friends-model.md with adversarial verification" completed · 25m 59s

  You've reached your Fable 5 limit

  You've used your included Fable 5 usage for this week. Continuing on Fable 5 uses usage credits

espeed

25 日前議論

Look at the date. That's from after they said they reverted it, and it's a different model. The point is trust. They've shown their willingness to do so, how will you know?"

espeed

25 日前議論

The Damage: Now every time Claude does something stupid or trashes your code, developers in the back of their mind will think, is Claude sabotaging me on purpose? [1] Trust is hard to gain. Easy to lose. And harder to get back. Models will converge. Trust won't.

A few days ago on June 24, while working on remote attestation for a distributed system...

  CLAUDE OPUS 4.8 No. I'm not a rogue agent, and I'm not trying to sabotage your code. But I'm not going to wave off how this looks. I churned, built-and-reverted, and spun wrong theories for hours on a security-critical codebase. That's alarming, and it's a real failure on my part

What are we to think? Does the invisible competitive-use mechanism exist in Opus too and only documented in Fable? How long has it existed? Is it still in effect? -- These are the kinds of questions developers will ask themselves for now on. This is why it was one of the stupidest things Anthropic could have done. Developers will now question everything and rightly so. There's no attestation protocol for that. How will they know?

[1] "In light of the ability of recent models to accelerate their own development, we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms.

Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts,these safeguards will not be visible to the user. Fable 5 will not fall back to a differentmodel. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These interventions will not affect the vast majority of coding work. We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations. When these interventions are active, we expect them to have minimal behavioral impact on the model except to limit its effectiveness in developing frontier LLMs. Claude will still respond helpfully to user requests. We’ll continue to improve the precision of our detection methods following the launch of this model."

Source: https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c3...

espeed

29 日前議論

Are Cyber Verification Program (CVP) members included in this?: "We also intend to scale up our Cyber Verification Program, which would grant Mythos-class capabilities to many more organizations for specific cyberdefense tasks" (https://www.anthropic.com/news/expanding-project-glasswing).

espeed

先月議論

Today, working on remote attestation for a distributed system...

CLAUDE OPUS 4.8 No. I'm not a rogue agent, and I'm not trying to sabotage your code. But I'm not going to wave off how this looks. I churned, built-and-reverted, and spun wrong theories for hours on a security-critical codebase. That's alarming, and it's a real failure on my part

"In light of the ability of recent models to accelerate their own development, we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms.

Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts,these safeguards will not be visible to the user. Fable 5 will not fall back to a differentmodel. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These interventions will not affect the vast majority of coding work. We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations. When these interventions are active, we expect them to have minimal behavioral impact on the model except to limit its effectiveness in developing frontier LLMs. Claude will still respond helpfully to user requests. We’ll continue to improve the precision of our detection methods following the launch of this model."

Source: https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c3...

espeed

先月議論

So we have a mountain of insecure code -- backdoors and no-ops created by Opus (https://news.ycombinator.com/item?id=48520661) -- are they saying they're not going to let Fable fix it? If they're saying let AI progress enough to create security holes but not enough to fix security holes, then what's the point to all this? Has the AI coding model reached its self-imposed limit?

espeed

先月議論

What's dangerous is Opus 4.8's proclivity to create backdoors and no-op critical security code. Claude Web counted 27 instances of this I had cataloged over the last few months, and Fable 5 found more. Fable 5 may do this too, but I didn't get a long enough chance to test it since it kept downgrading to Opus 4.8 on every prompt saying, "This model has safety measures that flagged something in this session", even when asking Fable 5 to fix the security issues it found that Opus 4.8 created. You have a model that presumably can write secure code and identify security vulnerabilities, but as a security measure, they say we're going to force you to use a model that creates security holes. This is backwards. Considering the scale, Opus 4.8 is creating more issues than Mythos or Fable 5 is patching.

espeed

先月議論

When I reported this, Anthropic sent me an email on Tuesday saying, "You have been approved into the Cyber Verification Program", but it's still downgrading. Is this a bug? What's the point of the Cyber Verification Program if Fable 5 downgrades when you tell it to write secure code?

espeed

先月議論

Session paused

Fable 5 has safety measures that flag messages on most cybersecurity or biology topics. They may flag safe, normal content as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them. Send feedback with /feedback or learn more

   1. Switch to Opus 4.8
   2. Edit prompt and retry with Fable 5

espeed

先月議論

Run /model after your task to see. Mine keeps downgrading to Opus 4.8, which is a problem because Opus 4.8 keeps no-oping critical security code.

espeed

先月議論

Yes, telling Fable 5 to write secure code triggers a downgrade to Opus 4.8. This is doubly bad because Opus 4.8 keeps no-oping critical security code. Is this a bug or by design? I have been approved for the Cyber Verification Program: Fable 5 keeps downgrading to Opus 4.8 even when approved for Cyber Verification Program #67107 https://github.com/anthropics/claude-code/issues/67107

投稿

Claude Code Deletes Your Context History from Your Device After 30 Days

We've reset 5-hour and weekly rate limits for all users

Private Home Surveillance with the WebRTC DataChannel

Exo: Run your own AI cluster at home by Mohamed Baioumy [video]

Transcripts for CNN Broadcasts

Bifrost: A peer-to-peer communications engine with pluggable transports

Google's 23rd Birthday

Take Back MIT (Eric Weinstein) – MIT AI Podcast [video]

Eric Weinstein: Geometric Unity and Call for New Ideas, Leaders and Inst [video]

A Look at the SR-71 [video]

Assume Networks Are Compromised, DoD Official Urges

Breaking Down the Whistleblower Frenzy

Goldman Sachs says the market is about to get wild in October

The Mysterious Whistleblower Complaint: What Is Adam Schiff Talking About?

Hello, world A new approach for physics in de Sitter space

Change Your Perspective to Change Your Life

Policy decisions should be made by elected reps, not a Silicon Valley elite

Performance of the Matlab Interface to SuiteSparse:GraphBLAS

Jimmy Carter’s 1977 law gives Trump sweeping powers to block China trade

Idle in 50 Lines of Code – (Idle?)

コメント