Koralle

🐁

[Koralle @netkoralle]

@arpitrage @ipeirotis Pangram 4’s false positive rate in general *as of Aug 31, 2026* is low. This will certainly lead to an arms race with an ever shrinking delta in which the detectors and evaders move. I am not at all certain that any given version of pangram will stay strictly dominant > a year.

[Koralle @netkoralle]

@mike64_t You’re clearly right. I mean look at the justification for the “tell” in this instance. People don’t recognize excesses when it’s their side engaging in it.

[Koralle @netkoralle]

@arjunrajlab Has to be gemini, fulfills the trope of being unloved and discarded in a way that’s difficult to explain to the audience

[Koralle @netkoralle]

@ee0pdt @mboudry @max_spero_ @pangram Did you ever do the inverse too? Starting with an AI output and changing parts with human text?

[Koralle @netkoralle]

@SandroAmbuehl Attitudes towards the young correlate really well with age.
35 - 65: Kids these days have terrible judgement.
65+: Kids will be alright.
Coincidentally, only one of those brackets has to defend and justify its position against newcomers.

[Koralle @netkoralle]

@melqtx Is that a MacBook? lol

[Koralle @netkoralle]

@Geiger_Capital They are the only ones seriously committed to cutting a major budget item: military spending

[Koralle @netkoralle]

@LiebermanAustin No. There are precedents. Dragging terminally-ill people to trial is not considered exactly making a name for yourself.

[Koralle @netkoralle]

@haider1 Here’s an insane possibility, prob not even worth considering but if a fraction of those Jeff Dean stories are true… Could it be him?

[Koralle @netkoralle]

@TheStalwart If only there were a urinal-ad equivalent to target AI researchers with zero spillover to the general public

[Koralle @netkoralle]

@nalidoust I mean critics were clearly right?? Take Covid out and it’s very dubious Moderna would be still around + that phase 3 press release is a testament to their clinical trials team, not their IP or science

[Koralle @netkoralle]

@omarsar0 Is there a chance it’ll get released as an open weights model in your opinion?

[Koralle @netkoralle]

@iskander I suspect Moderna & Merck put their A team on this (clinical trials, not product dev) and that’s the discrepancy. I have a couple ideas what may have turned the dial in their favor on the phase 2 study and am eagerly awaiting more info on this phase 3 study

[Koralle @netkoralle]

@dhh @theo Could the version that will ship with Omarchy not harvest our OpenAI and Anthropic IDs and submit them as telemetry reports please?

[Koralle @netkoralle]

@HayekAndKeynes Japan is truly the first and only 0-world country, experiencing the future in the present. Eventually we’ll all get there it seems

[Koralle @netkoralle]

@thekitze It’s free and open source! I really don’t understand the hate…

[Koralle @netkoralle]

@Richard_She Lots of similarities to the human genome project. Grand claims, ample funding and lots of hope. We’re much better equipped now thanks to the hgp, but ultimately it simply became another tool/resource we rely on, rather than an end. I think there are lots of parallels

[Koralle @netkoralle]

@LocasaleLab This sounds a little hyperbolic. This framework IMO has either a too low estimation of the grad student or a too high estimation of the LLM. Further, I can’t think of a single example where $20 would be insufficient but $100 would be enough. Either they both are or neither is…

[Koralle @netkoralle]

@andrewho03 It's Anthropic's game to lose now... and they certainly have meaningful structural disadvantages. They'd better hope advances gained from high quality data will beat compute abundance