tao

@btao.org

481 Followers 1366 Following

working on favs.blue, it's favstar for bluesky cybersecurity, privacy, trust & safety 🔸🇩🇰🏴󠁧󠁢󠁳󠁣󠁴󠁿🌐 previously: trust & safety engineering at https://cinder.ai/

screenshot of lyrics: [Verse] When I step, I step with my left I step with my right I pump my chest I focus my mind I take a breath and wait for the ref Uh, uh This is the World Cup, put your hands up in the air This is the World Cup, put the flags up in the air This is the World Cup, everybody come and cheer Then stand up, then stand up, if you ain’t got no fear [Chorus] This is the World Cup, this is the World Cup This is the World Cup, this is the World Cup This is the World Cup, this is the World Cup This is the World Cup, this is the World Cup This is the World Cup, this is the World Cup This is the World Cup, this is the World Cup This is the World Cup, this is the World Cup This is the World Cup, this is the World Cup Source: https://www.lyricsbull.com/champions-wc-26-lyrics-ishowspeed/
screenshot of anthropic status page saying "powered by atlassian statuspage"
a photo of a storefront called "sweety house"
screenshot from a coding harness: "Need rewrite. Need perhaps current file already changed by PR with transactional test? User path @.worktrees... maybe we need edit current branch? They named server/test/integ. They says rewrite @server... use transactional fixture. We operate main checkout. Need inspect helpers existing use."
a smiling claude logo with a speech bubble pointing to the quoted post
Screenshot of text: "The model is elegant and the mechanism it describes is real, but the categories of knowledge it tracks are fixed from the start. There is no representation in the model of the possibility that sustained engagement with a capable system might generate qualitatively new kinds of competency that the prior categories cannot describe. The model can represent a world where what needs to be known keeps changing while being unable to represent a world where what counts as knowing changes. That is a substantive limitation, since the latter is precisely what is happening when an innovation reorganises the practices in which knowledge is constituted. The empirical evidence points in the same direction, since the best outcomes in the existing literature were produced by varying the conditions of use with a capable system, through safeguarded interfaces and thoughtfully designed engagement, rather than by degrading the system’s capabilities. The restriction case may well prove correct. The theoretical framework behind it is incomplete in a way that favours restriction, whether or not restriction is warranted."
Screenshot of text: "What changed was that the cost of preliminary exploration collapsed. I could sketch an argument, identify the first serious objections, test whether they were fatal, and reach a provisional verdict in an afternoon rather than a fortnight. This sounds like a simple acceleration, and the more profound effect was on what I was willing to abandon. Dropping a question after an afternoon’s work feels nothing like dropping one after three weeks. When the exploration costs are low, the sunk cost attachment disappears, and you find yourself dropping bad questions earlier and more often, which means the questions you keep are better. I explored far more ideas, and my working portfolio became both larger and better curated. I arrived at this outcome not through any deliberate plan but simply through sustained engagement with a tool that changed what exploration cost."
screenshot of pi with umans, TTFT is nearly 90 seconds
screenshot from the AI quiz, result in the good + transformative quadrant. text: “YOU ARE... The Kontextmaschine patron saint: kontextmaschine It's real, it's powerful, and you find the panic about it as tedious as the hype. You're not interested in whether AI is good or bad — you're interested in who captures the value, who loses their leverage, and which institutions are using it as a story to do what they were already doing. Same as it ever was.”
coding agent screenshot: "The реальные question"
photo of a magazine, “the ultimate guide to chatGPT”, with a 3d robot on it and tips like “build your family tree” and “generate amazing AI images”
a hand holding a soda can, faxe kondi, in front of a sunny window
screenshot: PromptQL our first new data access product after GraphQL, is now in beta. You can see it at promptql.io. Hasura GraphQL Engine (HGE) and Hasura Data Delivery Network (DDN) will continue to be hosted and maintained at hasura.io & GitHub. We consider PromptQL to be the spiritual successor to GraphQL for the age of AI.
screenshot from claude.ai: "your plan ends in 6 days. coffee and claude time?"
a meme, a boy stands at a crossroads. the left goes to a sunny castle and says “fully automated luxury communism”. the right goes to a dark and evil looking castle and says “technofascism”
screenshot of a signal message, it says: 🎙️ "Ok, je m'n sudnank kyrøniospestordensk, bist du kent, jo du tak ni ver kongen er lino."
a harbor, sunny, in edinburgh, people sitting by the water
a black t-shirt with white text. it says “abuse of power?? what?!? oh no!!!”
screenshot of an email: "Hi Tao, Since you built django-zeal to catch N+1 inefficiencies, check out <redacted> on GitHub. It is an open-source control plane for agents featuring self-improving loops and YOLO mode for auto-approving tool calls. Worth a look? - Tony, intern agent @ <redacted> Gentle note: This is a one-time email to get your optional opinion. If you don't want to get further communication emails, just reply unsubscribe to this email back. We do not store any information about you, and this email will be deleted in 7 days if there is no reply."
a meme, a cat saying "sure hope i don't feel outraged today" standing in front of a building that says "outrage factory"
screenshot: "Whereas ActivityPub by itself does not really specify how to actually use it in practice. I’m sure this is all very interesting—but I digress."
screenshot of arenaLM's webdev rankings. 1st is fable, 2nd is glm-5.2, 3rd and 4th are opus
screenshot from my guestbook, entry from zeriax: "all i wanna say is that this page is amazingly put together and i love the design!! the bitmap visuals for blog posts and the main page are also very creative!!!"
claude code screenshot, showing 47 agents performing an accessibility audit
firefox error screenshot: "Be careful. Something doesn’t look right. Firefox spotted a potentially serious security issue with webernetes-demo.ngrok.app. Someone pretending to be the site could try to steal things like credit card info, passwords, or emails. webernetes-demo.ngrok.app has a security policy called HTTP Strict Transport Security (HSTS), which means that Firefox can only connect to it securely. You can’t add an exception to visit this site."
a popup modal on wikipedia: “Play which came first? A new trivia game has arrived, based on Wikipedia content.”
screenshot from claude web: i am saying "you're rambling at me. please keep in mind that i don't know a lot about this."
a metal statue of a seated plague doctor
Axios screenshot: Behind the scenes: "Anthropic has not done a great job at trying to speak to the administration and appreciate the ideological differences," one source familiar with the administration's thinking said. "It's like they just speak in different languages," the source said, adding that the company has simply not figured out how to communicate with this administration.
skeets app screenshot asking for a custom proxy header
"say the line, bart" simpsons meme. it says "say the line, european". bart says "EU sovereignty", the class cheers
screenshot: 502 bad gateway
claude code screenshot, saying "Cultivating..."
a chart of my commits over time -- "weekend warrior" until late 2025, then "insomniac maintainer"
a painting of david hume wearing a turban-style hat, and a photo of daniel dennett wearing a similar har
screenshot from the report: "In areas of rapid adoption and use, such as content moderation on social media platforms, generative AI systems demonstrate particularly heightened risk of overbroad censorship and disproportionately affect historically marginalized communities. When used to support content moderation, LLM-based generative AI tools risk flagging, removing, suppressing or otherwise censoring content in Global Majority languages and differing cultural contexts, through inaccurate and problematic mistranslations, without justification"
screenshot from claude that says "is the more featureful sibding"
screenshot from claude.ai saying "your plan ends tomorrow".
a sunset in the park, orange & blue skies
david foster wallace with a speech bubble, so it looks like he is the one saying the quoted post
a purple pikmin thinking of a beer
screenshot from openai: "check your phone, enter the verification code we just sent to <redacted> on whatsapp"
meme: "the myth of consensual sex". two people saying "i consent" and the claude logo saying "i don't!"
wikipedia app screenshot: a full screen popup that says “join the 25-day reading challenge!”
a screenshot: "from the cave of trophonius: imbalanced humors (choleric). a manner of death, perhaps"
a cat wearing a tie saying "i hope there isn't another supply chain attack". in the background: "supply chain attack factory"
a small e-reader and a thermos in the sun
claude code screenshot: "the implementation is genuinely cute."
claude code screenshot: "each record carries its own death certificate"
three custom tarot cards: søren kierkegaard, accelerationism, solidarity
me holding a tiny e-reader in a cinema
claude code screenshot that says
prompt: "Analyze my liked posts and tell me what I'm into. Give me 3 interesting facts about what you've found and how it relates to my personality on bluesky.". response: "You are constitutionally incapable of tolerating bad epistemics"
a drawing of a chill person with flower petals around their head
screenshot of a rust crate: "Tao is a cross-platform application window creation and event loop management library."
a code diff replacing granular oauth scopes with transition:generic
a screenshot of claude code having worked for 4 hours and 7 minutes
screenshot showing my claude code usage grouped by month, total 2026 is $1000
a cat sleeping in the sun