Will Oremus ✓

@WillOremus · joined 10 Jun 2010

Tech writer at @TheAtlantic. DM for Signal.

46 240Followers

3 963Following

6 430Posts total

376.1KViews on collected posts

Ultimi post

Will Oremus ✓@WillOremus3 Sep

Some interesting reporting here on Pangram and I don't think the reporter deserves to be piled on. That said, I kept waiting for actual concrete evidence of Pangram getting stuff wrong and it just... never really got there? 49.1K views · 129 likes · 2 reposts · 8 replies Open on X →

Will Oremus ✓@WillOremus2 Sep

On one hand, requiring interoperability seems like a not-unreasonable remedy to monopolization by a tech platform. On the other, where is the incentive in any of these decisions for tech platforms to avoid monopolizing in the future? https://t.co/sPrTAPlC5O 8.6K views · 29 likes · 7 reposts · 2 replies Open on X →

Will Oremus ✓@WillOremus27 Aug

The Wall Street Journal's opinion editor says the paper's subscribers don't need to know when an op-ed was written with AI. I predict that the verdict on this stance will come from his favorite source: the free market. My latest: https://t.co/QNcnaSQYab 22.7K views · 90 likes · 23 reposts · 8 replies Open on X →

Will Oremus ✓@WillOremus27 Aug

Personally I'm holding out for "Rogue" mode 1.9K views · 9 likes · 0 reposts · 1 replies Open on X →

Will Oremus ✓@WillOremus27 Aug

New: OpenAI is developing a "Persistent" mode for its AI agent, Codex. Persistent mode is a new level of reasoning effort that will make Codex "continue working until put to sleep." The agent will also proactively create follow up tasks for itself, per code reviewed by WIRED. ht 115K views · 561 likes · 53 reposts · 34 replies Open on X →

Will Oremus ✓@WillOremus26 Aug

randomly giving out genuine compliments is such an easy way to make the world a little nicer, i'm always surprised more people haven't figured that out and have a ton of respect for those who have 1.7K views · 24 likes · 1 reposts · 1 replies Open on X →

Will Oremus ✓@WillOremus13 Jul

@WillOremus This wouldn't explain why the behavior is absent in base models or fine tunes that eschew the style (I've made several that never do it). It's an RLHF artifact of some kind, not something inherent in token prediction 3.4K views · 34 likes · 0 reposts · 1 replies Open on X →

Will Oremus ✓@WillOremus13 Jul

Gift link to my story on why chatbots are always saying “It’s not just this, it’s that.” It’s not just an explainer—it’s a journey. https://t.co/v1Wtxqj7lE 21.8K views · 86 likes · 10 reposts · 4 replies Open on X →

Will Oremus ✓@WillOremus12 Jul

I set out to learn why AI models love negative parallelism (“It’s not X—it’s Y.”) This was the most interesting theory I heard: https://t.co/ieM1uTP2w5 151.9K views · 1.7K likes · 155 reposts · 29 replies Open on X →

Rispetto ad account della stessa dimensione

6 post degli ultimi 90 giorni, accanto alla fascia di 10K–100K follower. arriva a molti, ma pochi di loro reagiscono.

Visualizzazioni mediane15 664questo account924mediana per 10K–100K

Copertura, %33.88%questo account3.62%mediana per 10K–100K

Interazione, %0.55%questo account1.52%mediana per 10K–100K

Metrica	Questo account	Mediana per 10K–100K	Rapporto
Visualizzazioni mediane per post	15 664	924	16.9×
Copertura (visualizzazioni ÷ follower)	33.88%	3.62%	9.36×
Tasso di interazione	0.55%	1.52%	0.36×

Altri account di questa fascia → Confronta con un altro account → Come sono costruiti questi parametri →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.