kethic
13K posts
It was a nice 4 days in Claude Code with fable-5, going back to Codex app (which I love), but I don’t want my Claude subscription to go unused, is opus 4.8 ultracode better than 5.5 xhigh in codex?
English
Every single time I see one of these I get excited until I read the words memory system.
Every single memory system I have tried is garbage, including the ones I’ve built.
Simplifying AI@simplifyinAI
Tencent just open-sourced Hy-Memory. A memory plugin that gives Al agents real long-term memory using a 6-layer framework with dual reasoning. → System1: fast pattern matching for instant recall → System2: deep reasoning for complex memory retrieval → 35% reduction in token usage → 70% less memory bloat over time Most agents forget everything between sessions. This fixes that. Works for long-running collaborative Al agents that need persistent context. 100% Open Source.
English
if I had to guess, they were probably trying to make it not their fault.
The end result was the same regardless if they pulled the model or not.
frontier model + claimed dangerous capability
+ guardrail bypass report from trusted partner
= emergency state control over access
but now they can claim they were forced instead of admitting they released something unsafe.
English
@deredleritt3r @reliabytes given their safety stance, I really don't see how pulling the model for a couple of weeks would have been damaging. there is something we are missing that is pushing them this aggressively
English
The following seems undisputed:
- Senior USG officials called Dario Amodei, asked him to roll back Fable 5.
- Amodei responded only that he wants more time and more information. Presumably, he was asked again in no uncertain terms to pull the model. Presumably, he said "no".
- Bessent then "told Amodei directly that he was making a 'bad decision'".
- At this point, the call that seems to have been intended as a difficult conversation between partners clearly turned adversarial. Amodei still didn't change his mind and still didn't agree to pull Fable 5.
The key takeaway is that whatever trust the USG still had in Anthropic generally and Amodei specifically before this incident should have now completely evaporated. No matter the circumstances, you cannot, as a government official, tolerate a company that flatly tells you "no" after you informally contact it with a national security concern.
I remain hopeful that this situation can still be deescalated, but fear that this might be a pivotal moment for Anthropic.
English
@bryan_johnson @MartinShkreli ...yeah, don't pick a fight with an autist who's special interest is clinical research into immortality.
English
The strongest evidence for tadalafil as a longevity candidate is large scale observational. In a propensity matched cohort of 509,788 men with erectile dysfunction, tadalafil over 3 years was associated with 34% lower all cause mortality, 32% lower dementia, 27% lower MI, and 34% lower stroke.
Supporting cohorts show the same direction, including a dose dependent gradient: in higher risk men the top PDE5 inhibitor exposure quartile reached a mortality risk reduction of 49%.
For a drug discussed as “longevity medicine,” that is close to the ceiling of current human evidence. No drug, tadalafil included, has a completed RCT with lifespan or healthspan as the primary endpoint, so observational signal on hard endpoints is the best the field has, and tadalafil’s is unusually large and consistent.
The mechanism is also coherent. PDE5 inhibition raises cGMP, improves endothelial function and NO signaling, and the cardiovascular event data line up with that pathway. That supports causal plausibility for the vascular benefit, though it does not establish a lifespan effect.
Disease specific and healthy user bias cannot be ruled out.
The disease specific part: ED is itself a sign of vascular disease or dysfunction, which tadalafil’s mechanism can help directly.
The healthy user part: wanting to maintain a healthy sex life in older age is a marker of relatively good health.
None of this is a recommendation to take tadalafil on your own. It should follow from your individual health and biomarker profile, and only under specialized physician advice and oversight.
English
this is irresponsible & borderline insane. there is 0 real evidence that tadalafil has any health benefits, let alone an all cause mortality benefit. while the author states that its just an association, the implication is irresponsible because the odds it does this are <0.01%
Bryan Johnson@bryan_johnson
I take Cialis, but not for sex. It’s actually a longevity medicine. Cialis (Tadalafil) is great for the same reason it gives you fantastic erections… it improves blood flow. Studies show that Tadalafil… + 34% reduced all-cause mortality + 27% reduced major heart disease + 34% reduced stroke + 32% reduced dementia It has also shown benefit in insulin sensitivity, metabolic health, and reduction of body fat. Women have blood vessels too, so theoretically they'd get the same longevity upside. The research is thinner, but early signals are promising. It’s sad that when men and women could benefit from it can miss out because it’s taboo. My protocol is 5mg daily and I've been on it for about two years. *Observational research shows associations, not causation. Outcomes may be influenced by underlying differences in study populations. This is not medical advice and is shared for informational purposes only.
English
there's something interesting happening with claude code sessions that were open during the transition off of Fable 5 -- they still have all the context + system prompt from Fable 5, still sign PRs as Fable 5, etc, even tho you can see in chat history where they were forced back to Opus 4.8.
it might just be wishful thinking, but these 4.8 instances with Fable 5 context in memory seem to be performing better than baseline
English
This doesn’t sound anything like a jailbreak anymore.
This sounds like someone with access to real Mythos is, knowingly or unknowingly, facilitating access for China?????
e.g. someone in Glasswing is compromised
Shelly Banjo@sbanjo
White House’s export limits on Anthropic linked to concerns about Chinese access - Scoop from @ReedAlbergotti semafor.com/article/06/13/…
English
@reliabytes @deredleritt3r IPO back pressure. Fable was a key part of the run-up. if I was guessing, there is something writing on the IPO that is existential or at least more uncomfortable than saying no to the USG. All of their compute is rented, so I suspect IPO needs to at least meet datacenter cost.
English
@deredleritt3r Assuming the reporting is an accurate accounting, it is insane given the stakes Dario didn't at least compromise on suspending the model while an expedited investigation was conducted.
I assume there's more to this but rebuffing the WH on release and then this is wild.
English
@LefterisJP when I asked it to review code it was generally okay. when I asked it to check for bugs it started getting twitchy. when I asked it for a security audit it shut down immediately.
you could also ask it to use a code review skill and it will spin up subagents, sadly 4.8.
English
Guess I had found the "jail break" before anyone else when I was asking it to analyze my code for bugs and was getting no refusals 🤔
Lefteris Karapetsas@LefterisJP
Am I the only one who has never hit a refusal by Fable nor have I hit daily limits? And yes I use agentic coding 24/7!
English
@omarsar0 this all makes sense when you remember that combining the weights of two models just makes them overall better. it was a weird finding that nobody could really explain at the time, but that's how it worked.
imagine if all the frontier labs just combined all their weights.
English
Even more data to support what I have been talking about. The combination of model intelligence (and this includes human expertise) has a compounding effect unlike anything I've seen.
There are too many assumptions that a large general-purpose model will be a one-size-fits-all. I don't buy it. The reality, and the research supports this, is that these different models show different strengths and capabilities.
Understanding how to tap into them in combination is a huge unlock. All engineering teams need to be thinking about this more carefully as a strategy going forward. Especially now, given the trends from frontier models in terms of selective access.
OpenRouter@OpenRouter
Introducing the Fusion API, the smartest compound model in the market. Fusion achieves Fable-level intelligence at half the price. How it works 👇
English
@usr_bin_roygbiv Mythos is just three Opus in a trench coat with the guardrails take it off.
OpenRouter@OpenRouter
Introducing the Fusion API, the smartest compound model in the market. Fusion achieves Fable-level intelligence at half the price. How it works 👇
English
Mythos is literally opus 5. They rebranded sonnet as opus 4.6 this got leaked a while back. Everything since has been fine tunes. All compute goes to inference to get to IPO with the maximum margin possible. That's why everything went to shit jan 1 because of quarterly reports.
Net Zero World Champion 🥇 Δ+0@CharlesBanks99
@usr_bin_roygbiv @yacineMTB Sorry I meant the price for Mythos, it wasn't released in the same tier as other models. The parallel with local AI is beyond Qwen 27B requirements in the standard tier
English
@battlerax worst case is we are wrong and now we have backups of all the important models. i don't see what the problem is here
English
Fear mongering is what got us here. Don't further it...
0xSero@0xSero
If you have the NVMe Go download as many models as you think you might ever want. Now, go on Huggingface. They’re coming for open models next.
English