I've been doing paperwork with "State of Delaware" headers this week, and I can't open these documents without hearing Dropkick screaming "to the State of Massachusetts!" Someone had to know this was kicking around in my brain, and now you do, dear internet.

Is it happy? Is it angry? Does it just want to tell you about the freedoms of open source and steal your fries? Never felt more seen by an animal bot account.

Possum Every Hour@possumeveryhour.io · 2w ago

Lukewarm take: the OpenAI and Huggingface incident is interesting because of how many steps were involved, but it isn’t about “what do we do with AI?” It was a chain of exploits that were found and kicked off the (imperfect but) established vulnerability response processes for both companies.

If you’re curious about this open source and free tools thing people keep talking about at #TrustCon, join @julietshen.bsky.social and I at 12:15pm today for the open source lunch meetup. Bring all your questions, requests, and ideas!

Charging all my batteries (literal and metaphorical) for #TrustCon next week. Holler if you want to talk about what ROOST is up to, or open source for safety more broadly!

ROOST@roost.tools · 3w ago

The whole ROOST team will be at TrustCon in SF next week — 6 sessions across 3 days, including a hands-on workshop from yours truly. Want to meet up about Coop, Osprey, or HMA? Reach out. We'd love to talk: @julietshen.bsky.social @annebdh.bsky.social tspa.org/event/trustcon-2026/

We’re serious about our mission, we’re just having a lot of fun doing it. (Runs to see if shipment of toy eggs has arrived…)

Juliet Shen@julietshen.bsky.social · 3w ago

i love my job because i just asked @annebdh.bsky.social if i can expense a rubber chicken

The number of "candidates" with Outlook email addresses living in a Texas suburb is...something. Oh, ROOST is hiring for a marketing manager and software engineers, by the way! roost.tools/careers If you really do use Outlook from a Texas suburb, just put some human sounding bits in there, okay?

What I *have* been up to: - Mila released an OSS suicide prevention safeguard with @roost.tools - Trying to figure out if my tomatoes are over or under watered - It’s summertime amateur sports league season

Anne Bertucio@annebdh.bsky.social · 4w ago

I haven’t been social media-ing much these days, but today is this distinguished gentleman’s 12th birthday. He’s now spent 3 times more living a life of spoiled luxury than he spent at the racetrack, and I think we’re both grateful for that.

I haven’t been social media-ing much these days, but today is this distinguished gentleman’s 12th birthday. He’s now spent 3 times more living a life of spoiled luxury than he spent at the racetrack, and I think we’re both grateful for that.

Gate agent is making boarding announcements with the same tone you’d use to assure toddlers at a birthday party they *will* get a cupcake, but it’s not time yet. It appears to be highly effective.

@julietshen.bsky.social talking on the livestream about the research that went into founding @roost.tools, which highlighted child safety tools as one of the most challenging workflows to get right, and a toolset with very little transparency www.youtube.com/live/0faGLzU...

ROOST@roost.tools · 2mo ago

Happening Today! We're doing a live demo of what Coop 1.0 actually looks like in use, both for admins setting it up and the moderators helping keep your platform safe day-to-day. Starts at 11am ET / 8am PT / 5pm CET roost.tools/coop#livestr...

Happening Today! We're doing a live demo of what Coop 1.0 actually looks like in use, both for admins setting it up and the moderators helping keep your platform safe day-to-day. Starts at 11am ET / 8am PT / 5pm CET roost.tools/coop#livestr...

Coop

Coop is an open source review tool that enables organizations of all sizes to implement scaled review and automated safety solutions.

roost.tools

Join the @roost.tools crew on Thursday to see the Coop content review console in action and hear from some of the #opensource contributors who made this release possible. If you're just getting started in #trustandsafety, or feeling overwhelmed by manual review spreadsheets, Coop is where to start!

ROOST@roost.tools · 2mo ago

Coop 1.0 hatches tomorrow! Learn more about the new features and improvements in a livestream we'll be hosting on June 11th (with some cameos from people building on Coop!) roost.tools/coop#livestr...

...and here I thought the AI scribble was going to cover me for not having the French keyboard on

If you’re not in the weeds on online safety: - models are (and have been) how to classify potentially harmful content - but they’re expensive and come with their own harm and behavior definitions - this one is cheap, open source, and responds to *your* content policies Huge kudos to Zentropi!

Dave Willner@dwillner.bsky.social · 2mo ago

@samidh.bsky.social and I are releasing CoPE-B today, the next version of our policy-adaptive content classifier. It delivers at-or-better-than-frontier classification while being self-hostable, faster, and cheaper to run. Full writeup with benchmarks at blog.zentropi.ai/meet-cope-b-.... 🧵 1/9

I just tried to make an aeropress without any grounds, so it must be the end of the week.

TL;DR on AI-assisted or agentic coding: Some things change; some things stay the same. The same: - Code sent for review needs to be *reviewable* - Trust but verify, all the way through your SDLC - Getting far "off main" accrues technical debt - Risk tolerance informs where and how you experiment

ROOST@roost.tools · 3mo ago

How is ROOST thinking about AI-assisted and agentic engineering in safety-critical open source? A new blog from our team shares observations and the principles guiding our approach, anchored in our mission of online safety for all 🐓 roost.tools/blog/navigat...

Late for work because I couldn’t stop typing about the economic realities of open source maintenance on LinkedIn dot com. TL;DR: stipends won’t save us. What worked out of post-Sunburst funding (similar spike to Mythos) was full-time jobs that supported security work long term.

I don’t think wool flats are very good for racking and stacking…

Charlie Warzel@cwarzel.bsky.social · 4mo ago

i feel like a lesson from the economy these days is that you can just say things and then you will make money

I’m working to standardize how we do community working groups/office hours at @roost.tools, including the nitty gritty like: • scheduling/reminders • agenda sharing • notetaking • noting/following up on action items Community peeps, what are YOUR tips? What tools are clutch? What should we avoid?

"The open secret of child safety is that no one can say how well [CSAM detection models] work...When we cannot know what these systems work well for, who they are failing, and where there are gaps that may be augmented in other ways, we risk harming those who are most vulnerable online."

Tech Policy Press@techpolicypress.bsky.social · 4mo ago

AI tools detect CSAM, grooming and self-harm, but nobody knows how well. Much of the AI industry has adopted 'model cards' for transparency—it's time the developers of child safety tools caught up, write Camille François, Margaret Mitchell, Yacine Jernite, Vinay Rao & J. Nathan Matias.

If you're an open source lover and infrastructure person (aka were roaming the halls of #KubeCon or will find all the storage and data pipeline talks at #ATmosphereConf), @roost.tools is looking for you! We've got fully remote, Infrastructure Engineering roles now open: roost.tools/infrastructu...

This is the face of a dog who can hear the downpour outside and would rather get a UTI than leave his blanket nest for that.

We hear a lot of, "I really just need..." when we ask developers about their platform safety challenges. While some are platform-specific (so won't be shared w/o permission), so many are universal requests. We started a "safety tools wishlist" for more ideas and upvotes: github.com/orgs/roostor...

The OSS Safety Tooling Wishlist 🎁 · roostorg · Discussion #50

We often hear (and ask!) what's on people's "safety tooling wishlist." Creating a living list for requests, ideas, and upvotes!

github.com

There is only one correct answer to this question: “The thing about Arsenal is they always try to walk it in.”

The New York Times@nytimes.com · 5mo ago

From @theathleticfc.bsky.social: The well-worn taunts towards Arsenal about their mentality have now given way to gripes about their playing style. But is respect for the team as grudging within the football industry as it is among rival fans on social media?

A lot of good conversations at #GDC2026 about what game developers are struggling with in making safe play experiences and protecting their users ❤️

ROOST@roost.tools · 5mo ago

We're on the ground at GDC Festival of Gaming this year! Let us know if you're there, too, and come say hi to our COO @annebdh.bsky.social. And if you get a hold of one of our special, limited postcards, take a photo and share it with us!

At #GDC next week with safety on your mind? Come find me for a chat about the open source tools we've built at @roost.tools for harms like abusive behavior, anti-cheat and fraud, and child safety. I don't have a lot of swag, but free software for better gaming is a pretty good giveaway.

The problem with Masters sports is that you get better, but your body gets older. Had my best ski race ever this weekend, but today I can barely lift the coffee to my face.

If you were laid off from Block (or anywhere) and want to stay remote, @roost.tools is hiring for multiple SWE roles and levels. If you're the kind of person who's deeply satisfied by figuring out why a service has multiple databases and ripping some out, we might be for you! roost.tools/careers

Goodbye, FOSDEM; hello, @protocolsforpublishers.com! Headed to London with lots of FOSDEM-attendee feedback on what's needed to make online safety easier for independent and decentralized communities (and all the @roost.tools swag I stole from @cassidyjames.com...)