Try as I might, 2 kids in an urban arrow with e-assist off in Cambridge does not get me to cville sweat levels

If you downsample the image identically to what opus 4.6 would have done, is the image token usage equal between models? Feels way too linear that you see a 3x token increase when they increased the allowable resolution 3x.

It’s a big jump but 1) they are advertising improved image handling with 4.7 (the resolution increase is also 3x, probably not coincidence), 2) are these really price “hikes” when it’s for a new model, and the old model is still available?

Just sent a request. Will share updates here.

The revised list of request parameters would be very easy to fill out. I’d be willing to do so independently as another gauge for whether the authors are attempting to hide incorrect findings or have legitimate consent/commercial reasons for the process they put in place.

The study coordinators may in fact have procedures in place (putting aside, again, whether they were reasonable to establish in the first place) that do need these written in plain words (not everyone involved in human study conduct is immersed in open science discourse…)

Prefacing this by saying I don’t think a reasonable patient consent process for this study would warrant this degree of restriction.

But did you follow up as requested with “study objectives, research hypotheses, detailed methods, preliminary data, sample size estimation, and proposed conclusions”?

Bonus: with these mirrors, your feed doesn't get populated by anthropic's gracious replies to "tHeY NErFeD dA MoDEl" or "what do you mean I can't deploy Claude as a service in my app using my $100/month subscription?" engagement bait.

The Anthropic team does lots of helpful posting to X, but unfortunately has not prioritized bluesky cross-posting (makes sense, X algo loves everything AI)

@j4ck.xyz made this tool to mirror those accounts, and takes community requests.

Boris @bcherny-x.bsky.social
Thariq @trq212-x.bsky.social

Thank you for your service—sent in a couple requests!

Trying to reduce my use of social media with algorithmic feeds, and found this very nice bot today.

So much of Anthropic’s dev and media teams just post to X as a sole outlet. 2nd best is subscribing to new release notes for Claude code on github.

But effort level selection is still available in Claude Code and the Claude API.

More discussion on Twitter threads. Adaptive thinking seems to be more considerate of when to enter extended thinking, reducing token usage (but also no longer allowing users to explicitly toggle thinking effort level)

I never saw issues w/ Opus 4.6 but did with much earlier models, so gave this a try:

“Give me media conditions for growing C. difficile for bacterial metabolism studies”

Opus 4.7 ❎
Opus 4.6 ✅
Sonnet 4.6 ✅

I do wonder whether it is related to memory that doesn’t xfer across models.

5 years in to corporate WSL use, I am back to believing there is too much friction, and potential for organization-level IT policies to interact unfavorably, for more than sporadic use. Academic institutes with more flexible IT, or cases where you can spec the hell out of your computer = exceptions.

Very weird and unfortunate but their interpretation makes sense, seems like the intent is only to cover purely educational activities and require an additional license purchase for non-profit research activities.

All of our extra capacity is parking lots. Even the regular trains to the stadium are “event” trains: www.mbta.com/destinations...

Regular service doesn’t have the capacity—they are running extra trains

The whole thing is hilarious to me. In Boston, people are surprised that commissioning a special train, just for their tournament (that people are flying in globally for), is more expensive than the commuter train that has operated 7 days/week for decades on the same schedule.

@ec.europa.eu DG HERA announced a €30 million investment managed by KfW, Germany’s Development Bank, to CARB-X and @gardp.bsky.social to strengthen the global response to AMR. CARB-X will fund early antibacterial research, while GARDP will advance clinical development. #IDSky 🧪
🔗 bit.ly/4v5K81M

Sorry that happened. Amazingly, there wasn’t any category I could report the post under 🤷🏻‍♂️

Hey Peter, completely unrelated, but your Twitter account looks to be hacked (you have a new pyramid scheme post about bitcoin and a Mercedes SUV…)

Press release: www.businesswire.com/news/home/20...

Vedanta Biosciences Announces the Phase 3 RESTORATiVE303 Study of VE303 for the Prevention of Recurrent C. difficile Infection Will Continue as Planned Following Protocol-Specified Interim Analysis

This is wild, we eat so much yeast

Important for people just skimming this: the consortium is all enterococci.

Yeah this is not a surcharge on commuter rail, this is special service… I want my taxes going towards better regular service first, not expenses for soccer tourism.

Her son needed a fecal transplant for a fearsome C. diff infection. Getting one required a tortuous journey Her son needed a fecal transplant for a fearsome C. diff infection. Getting one required a tortuous journey.

C. difficile is a terrible condition. Seems like it shouldn't have to be this hard to try to treat a child with a bad case of it. www.statnews.com/2026/03/23/c...

Bacterial Consortium Therapy for Prevention of Recurrent C difficile Infection This clinical trial compares the efficacy of 2 different doses of VE303, a defined bacterial consortium of 8 strains of commensal Clostridia, and placebo in preventing recurrence of Clostridioides dif...

I see your drug with 4 active ingredients, but introduce you to ours (also for recurrent C diff infection) containing 8 bacterial strains. Also no fecal matter needed: jamanetwork.com/journals/jam...

Novo Nordisk Fellowship Potential areas of focus for postdoctoral fellows participating in the MIT-Novo Nordisk Postdoc Program include, but are not limited to, the following: Your complete application should include: In the...

If you're a senior PhD student or recent graduate interested in doing work at the intersection of AI and microbial genomics and/or microbiomes, please feel free to reach out to discuss the possibility of applying together for an MIT Novo Nordisk Fellowship engineering.mit.edu/novo-nordisk