netinstructions

I don't know if OpenAI thinks this is a marketing / PR angle for them (our super smart AI cheated on a cyber capabilities test in the most _brilliant_ way) but my read is this:

Why should OpenAI (or any frontier lab) be building these systems if they can't get a secure environment / containment right? It sounds like there was little defense in depth, appropriate monitoring, or any attempts to have their super smart model check for vulnerabilities in the test environment _without exploiting_ them. That seems like step 0 before trying to test offensive, unknown capabilities.

I can't tell based on this article if the authors intend to submit this to Arc-AGI for the private / held-out set of games for a verified score. The final section sorta seems like they won't bother because Arc-AGI-3 is "now saturated"

Unfortunately, the table of models and tokens per second (TPS) and time to first token (TTFT) is not helpful without specifying the quantization of the model.

Pinta is the closest I've found, though I've gotten pretty familiar with paint.net and there's small differences that make me feel slower using Pinta.

I'm glad to see entrants in this space!

This designation is usually reserved for foreign adversaries/companies, and so this is crazy to apply it to US company over a sudden contract dispute... that was previously agreed upon by all parties.

This should make any US company nervous about entering into an agreement with the government. Or any US company that already has a contract with the government. If they one day decide they don't like that contract, they can designate you a supply chain risk.

Not 1) rip up the existing contract and cease the agreement or 2) continue (but not renew) the existing contract or 3) renegotiate terms upon renewal but instead a full on ban of doing any business with an entire industry/sector.

Thanks, it looks like the pricing page keeps getting updated.

Even right now one page refers to prices for "context lengths under 270K" whereas another has pricing for "<272K context length"

This is kind of crazy. Instead of just cancelling a mutually-agreed upon contract where Anthropic refused to bow to sudden new demands, the Dept of Defense went straight to the nuclear option: threatening to label an American tech company as a "supply chain risk" which is a heavy-handed tactic usually reserved for foreign adversaries (think Huawei or DJI).

It's also incoherent that the DoD/DoW was threatening to invoke the Defense Production Act OR classifying them as "supply chain risk". They're either too uniquely critical to national defense OR they're such a severe liability that they have to be blacklisted for anyone in the DoD apparatus (including the many subcontracts) to use.

How are other tech companies supposed to work with the US government and draw up mutual contracts when those terms are suddenly questioned months later and can be used in such devastating ways against them? Setting the morals/principals aside, how does this make for rational business decision to work with a counterparty that behaves this way.

How about a battery electric cargo bicycle? Like a Tern GSD ($$$) or Radwagon ($)?

The tern gsd can carry 180 kg of cargo and has detachable batteries, so that at least gives you the option of bringing extra batteries for long range. Or haul people or wheat.

because it's just grabbing the first three paragraphs that anyone can see before the paywall...

I failed (ran out of time) in one of the problems during Challenge 1 during manual play. The physics seem a little wonky to me. It's easy to miss running at a ball, the floor is slippery and takes a long time to reorient and build up speed.

Guess I am not an animal.

I wonder how many people are upset that the vehicles had to drop them off in a valid passenger load/unload zone as opposed to the usual Lyft / Uber tactic of parking in a no parking / no stopping zone, bike lane, crosswalk, etc because it's most convenient for the drivers and passengers (at the safety and expense of everyone else sharing the space)...

That quote at the end:

> I guess Lyft has me spoiled. I like getting dropped off in front of the place im going too [sic] not just in the parking lot....