6thbit

Likely just trying to appear relevant in the news cycle.

It's also kinda wild how the competition being at v6.1 makes 3.x feel ancient, at least saying you are at v4 now changes public perception a bit imo.

Have each agent rewrite openssl in $lang and call it the RollYourOwnCrypto bench.

I wonder if there's people already whose full time job is maintaining/extending one of these auto-migrated codebases.

Imagine they aren't even familiar with rust but are deeply familiar with the product.

  > forecasts that annual spending on AI infrastructure might hit $1.5 trillion by 2031. 
  > ...
  > sustaining this level of investment would require an AI market approaching $6 trillion annually

This is a 2031 projection of a 6T market, but somehow the article editorializes this into 'justifying' a boom?

I had forgotten OAI bought openclaw this year. Are dots are what came out of buying that team?

Quite an upsell from $20 to $100 to get dots, especially in non-US markets.

On the other hand, Meta is not making money from muse base tier yet.

So, if OAI finds a way to make money the same way meta would for their free tier, maybe they follow suit.

I'd like a brainstorming mode, with limited side effects.

I want it to go and read stuff but also invoke some commands and generate reports and discuss on the results. Latest GPTs in codex plan mode always go.. write a plan. Shocking, i know, but thats not what i want on every turn when im in that mode.

Fair point, and i think that's fine. AI labs could just not jump themselves into such problems and let humans have a go at them, whether ai-assisted or not. As in this case with a regular budget one researcher could've access to.

It'd be hilarious that meta chose to pay openai. If there was such a deal, would it come to light in any public/official filings?

  > As it turned out, the result wasn’t all that far away for humans either. A few days after I heard from Anthropic, we heard from Song He, an amplitudeologist at the Chinese Academy of Sciences in Beijing. Song’s group had already gotten the majority of the result. They’d used some AI assistance, based on GPT-6, but not the kind of one-shot almost human-less approach Anthropic used.

Please have AI come up with something no human is also about to solve?

This gets me wondering why ai labs aren't proposing their own millenium prize type challenges.

Huh, didn't know the font could control variations over nth glyhph!

Could one make in theory a regular monospaced font to just render glyphs past 80 chars of a line as bad as possible?

There must be Pro Sailing kits already under the price of a few of these things.

Any good starting point for a fisherman not interested in stashing fish but just capturing and releasing?

They only anthropomorphize agents when convenient.

lol, it's true, Claude in the article is so anthropomorphized that you could read it as if it was a human just as well.

It's really a matter of iterating on the problem and validation right?

Models make progress on coding and math because they can write tests and proofs to an extent. Many industries that are more 'physical' and require performing experiments lack that instant feedback loop. Find a way to close that loop and AI begins to look useful.

But try and convince companies to invest on closing that loop just to see if the current models work well on their problems or not? Tough sell. So Anthropic just shows them, hey look, this is possible and if you don't do it I will.. so they fold.

The article talks about the lab, so it's not quiet at all.

Lands as an active threat. Maybe they're serious about this research or not, but for sure medical companies doing this sort of research will consider upping their AI budget and connecting their labs, etc. to avoid "falling behind".

So, the AI labs benefit either from achieving something they could market or from the peer-pressure imposed to companies in the sectors they get their nose in.

when they say "claude" what model do they mean ?

no mention of opus/mythos/fable or anything..

Why are there results for opus5.5 on arc-agi 1 and 2 but not 3?

https://arcprize.org/results/anthropic-claude-opus-5-5

wait what debacle?

( why is the x-axis on the first chart in descending order ? )

since it was sso for their codex/chatgpt account, presumably they used an existing chatgpt github connector available to such accounts.

Somehow the auth'd token for the forum was available within that VM, so having RCE they could've replayed it? idk

  > but the commit was not documented as a security fix and received no CVE.

There must be an entire class of open source commits that unknowingly fixed security bugs without being tagged as security fixes that one could look for missed backports. Scary.

"quick" tunnels you can make by installing cloudflare software and running a command, without a cloudflare account, using a fixed "trycloudflare.com" domain and not your own domain.

existing (slow) tunnels you can create once you've set up your own domain for cloudflare to manage its DNS, installing cloudflare software, logging in to your account, and running a similar command.

edit: my bad, quick ones aren't new at all

The things in the 'what should be self-driving' section are the exact things i give to juniors so they can build their mental model of the codebase.

Now, I agree it'd be fantastic to have that automated, but how do we then share that context with the humans that will drive the non-self-driving actions?

So there's a strategy shift here. They launched financial services specific tools and now law?

Is the play here a set of specialized harnesses using their best general model?

  > the Tanstack compromise is very likely to have been the leak vector
  ....appears to have been backdoored to extract an API key with authorization to read the private codebase.
  ...
  >  immediately rotated all required tokens & credentials to prevent further incidents.

Rotating the API key doesn't quite put them in a position to "prevent further incidents" does it? The next PyPI/npm supply chain issue will just get the new key?

I suppose whatever they use that key for should be reviewed and re-scoped if possible?

Does github let you restrict where you can originate requests using a given API key? or are we just not there yet?

In my experience, Luna is not able to find non-trivial impacts from the changes at hand.

I've seen this a few times on relatively simple changes on complex codebases.

If they intend to keep training then they can have, privately, increasingly advanced models that never become public because they don’t meet their “pacing requirements”.

Then they all build a larger moat cause china doesn’t get to distill private models for a while.

As long as the consumers “buy” this pacing and keep paying for current level public models, doesn’t seem risky from a business standpoint.

Alas, that’d also leave us gov in a position to seize the private models at any time.

This is so ridiculous that any bad actor could use “OpenAI agents” as a plausible cover for anything today