novaleaf

Reminds me of the gain-of-function, COVID lab leak hypothesis. It seems like humanity just can't stay away from Pandora's box.

frontier vs "not quite" :D

makes me wonder if the starbucks story is fiction too :P

no, I didn't get access to sol until a few hours ago. I just have my claude protocol files linked to inside codex. trying Sol this morning for the first time, so I can't really comment on that vs gpt5.5.

However you can do what you are asking "fable--> sol" you need to setup a mcp or have fable run a bash tool, just invoke the `codex.exe` cli tool with whatever cmdline args are needed.

I sub both codex and claude at 20x. I like opus+fable more than gpt5.5 because it seems gpt tries to finish tasks by leaving any ambiguity unresolved. claude seems better at surfacing open questions.

This is using the same AGENTS.md prompts, which were designed firstly for Claude use, so maybe it's something that could be optimized better if I understood gpt as well?

I think the story was Starbucks -> Seattle's Best Coffee, not McD's --> BK. but it does work I guess.

reminds me of SBC's (Seattle's Best Coffee) strategy, which was decidedly not Nash: put a store across the street from every Starbucks.

it works by ppl harvesting low ranked submissions and resubmitting them via a (unofficial) voting ring.

I used to live in Thailand, and over the last 10 or so years it seems that the kind of DIY content you find on Youtube has had a positive effect. Before then it was difficult to find a building maintenance team that understood the purpose of a waste water P-Trap (unfortunately I'm not exaggerating)

yeah, for example, just send a hash of the domain used. but then maybe people would say anthropic is spying on everyone, instead of targeted spying...

obligatory link: https://en.wikipedia.org/wiki/Parallel_construction

it was released a few hrs ago as "Fable 5". it's an incremental improvement over Opus 4.8.

just yesterday I felt that claude code was being aggressive in it's defense, so I lead my response with "Spicy Take! Here's why I think the bug is happening...."

Because of syncopathy it took my "Spicy Take" and decided to say basically "Even more than it could, your bug is happening RIGHT NOW"... which was just made up lies for dramatic fit.

Back to talking to Claude like I'm a robot I guess.

the xAi revenue comes from renting their infrastructure to their biggest competitor. Anthropic is going to IPO for aprox 1/2 the valuation, is profitable, and can cancel the contract with 90 days notice.

I like to joke that if you look at every Feng Shui rule through the lens of "to reduce the risk of assassination" it all really makes sense.

Maybe it's not so much of a joke....

Max 20x is for individuals only. (could probably have emps get it themselves, and reimburse)

I think many people who wouldn't be disincentivized by demerits/expulsion/etc would find the threat of intense physical pain highly motivating.

Caning is no joke.

I must have read too many british books as a kid, because that's the way I use it.

Though, they don't mention the "I am sympathetic to your plight" version of sorry that seems to confound ppl a lot (which I also do)

I (and many others) are the opposite. I run out of quota is 4-5 days. Generally no issues with the 5hr cap. ($200 sub)

yeah, this smells like a bug in their (dumb) usage segmentation.

For example, there is a distinction of what is classified as extra-usage-billed VS extra-usage-enabled. As a long time claude user, I can assure you they are different things: to use Sonnet[1m] you are required to have extra-usage enabled, but it won't actually bill it unless you are out of quota. Surprisingly, you can use Opus[1m] without extra-usage enabled (!!!).

I think it would be a considerable improvement to allow duck-typing over the top. implicitly defined interface that includes exact member matches, something like that.

in Claude Code's conversation transcripts it stores messages from subagents as type="user". I always thought this was odd, and I guess this is the consequence of going all-in on vibing.

There are some other metafields like isSidechain=true and/or type="tool_result" that are technically enough to distinguish actual user vs subagent messages, though evidently not enough of a hint for claude itself.

Source: I'm writing a wrapper for Claude Code so am dealing with this stuff directly.

yes, I think his book "Reamde" had the concept of "Ameristan" which was the redneck-Idaho equivalent of many racial/religious/socioecon tropes.

A prosecutor would have to prove that you have knowingly and wilfully committed the crime in order to be convicted

IANAL also, but counterpoint: https://www.goodreads.com/book/show/6611240-three-felonies-a...

your linked bug is a cherry pick of the worst case scenario for the first request after a resume.

While it should be fixed, this isn't the same usage issue everyone is complaining about.

I'm sure this is no surprise to anyone who has used CC for a while. This is the source of so many bugs. I would say "open bugs" but Anthropic auto-closes bugs that don't have movement on them in like 60 days.

crazy solution that might work for you: open an incognito browser and check for deals for new customers. "someone" I know was able to switch from a $50/1.2TB limited 300mbit plan to a $45/unlimited 1Gbit plan doing this.

if they have a better deal for new users: sign up for a new account under someone else in your household, and cancel your old account after you get your new account hardware setup and working.

The 66-year-old brick will have the exact same interference fit, the same clutch power, the same 4.8mm stud diameter.

Pretty sure this is false. Old bricks had way higher clutch power, so high that it was deemed too difficult to separate. Sometime in the 90's the grip strength was reduced.

This false claim is underpins the entire article :(

you are replying to a bot, that's why.

genuine question: how would you turn a debugger-mcp into a CLI? do CLI tools used in agents have the concept of persistence?