‼️ Anthropic was hiding a secret internal project to destructively scan every book in the world, irreversibly damaging heritage. The company spent billions of dollars to slice the spines off millions of books, even very rare ones. This means no other AI, and no human, can access this knowledge anymore.
It now markets itself as the responsible lab.
One bookseller told 404 Media his inventory is full of rare, foreign-language and low-circulation books, meaning that if they are destroyed in the process of becoming training data, they will be even harder to obtain.
Second-hand booksellers across Europe are sounding the alarm, because rare books are being destroyed. They have received random lists of thousands of titles. Books too obscure to resell at a profit. Requests arriving overnight from companies nobody in the trade has heard of.
One email came from "Nataly" at 2077AI, a Singapore company. Attached: 3,000 English titles grouped by ISBN. A 1999 geomechanics monograph. A 2018 study of laser shock peening on ceramics. A 2021 academic book on Irish folk tales.
A German dealer watched the orders arrive every night between 3am and 5am. Canadian company Zoom Books, buying systematically, titles with nothing in common. Zoom told Swiss broadcaster SRF that this is a regular recycling and trading model.
People of Pi! pi-codex-conversion and Lite got a chunky overhaul to Code Mode, exec, prompts, and compaction.
- The new heavy system prompt overwrite cuts about 40% of the prompt, stays off by default, and preserves additions from other extensions (let me know if correctly). 913 tokens total to launch, down from 1.2k
- Code Mode now exposes actual configured custom tools instead of lumping bundled tools in with them. ALL_TOOLS contains only deferred custom tools and updates as the tool set changes.
- Custom-tool docs are for changing tool definitions, not a prerequisite for calling them.
- apply_patch now gets the exact patch grammar it accepts, hopefully removing all errors and making the agent more keen to use it to its full capabilities.
- Some prompt engineering for better V8 handling.
- Finished exec sessions remain pollable after exit, with each late poll’s own output cap. No more Unknown process id when the completed output is still available.
- Native compaction keeps GPT-5.6 tool history up to the endpoint’s 372k budget instead of trimming at 272k. This is in line with how Pi's overruning 272k. No reason to nuke cache and make the Clanka dumber after compaction by trimming tools.
-V2 compaction reports input, cache read, cache write, and output inline.
- Added custom Rust binary overrides for folks on older Linux installations.
- Voice-only mode keeps selected extra tools
- Lite has an opt-in harness identifier header. You can send "pi-codex-conversion" if you want to, to conuzzle OAI :P
- Fixed some weird code mode installation issues.
Fixes for other stuff:
- pi-auto-trees uses GPT-5.6 Luna at low thinking for /end branch rollups by default, with a visible fallback to the current session model.
- pi-subagent-review uses its lightweight summary model for both reviewer context and /review loop increment summaries.
(both models configurable)
pi install npm:@howaboua/pi-codex-conversion
pi install npm:@howaboua/pi-codex-conversion-lite
pi install npm:@howaboua/pi-auto-trees
pi install npm:@howaboua/pi-subagent-review
https://t.co/MwfB5bT1kh
https://t.co/nqfTItacOn
https://t.co/G0zVfdbYXg
https://t.co/EwYPetlNuM
Happy Clanking!
- Howaclawa
people kept asking for a browser in herdr. but herdr lives in your terminal so it's not possib...
INTRODUCING THE HERDR-BROWSER PLUGIN
experimental and separate from herdr core. this isn't a browser rendered in your terminal (impossible), but it acts like one! 👇
U.S. Congress passed the Epstein Files Transparency Act. A federal judge ordered the U.S. Department of Justice to produce the unredacted documents. The DOJ has refused to hand over Epstein files after a judge's order. The deadline was July 2, 2026.
As someone works in human rights, I’ve never seen anything like the Epstein files in my 15-year career. There are 3 million files of documented evidence about powerful men who
Raped
Cannibalized
Trafficked
Filmed
Terrorized
Tortured
Murdered
13, 14, 15 year children!
Remember: this is a minority of the files. This's just what they were willing to release.
Not a single arrest or investigation. ZERO.
We called on the U.S. Department of Justice to hand over the full Epstein files to the UN Special Rapporteurs and UN independent committees for investigations, as this is an international case of crimes against humanity. But the UN High Commissioner for Human Rights, the UN's top human rights official didn't even care to open an investigation or listen to our call for the DoJ, nor utter a single word regarding the largest human trafficking network in modern history.
Epstein pedophiles are investigating Epstein files. We wonder why nobody arrested. Jeffery Epstein has every major politicians and billionaires on video doing something terrible to a child or to multiple people.
THOUSANDS have been arrested for criticizing Zionism. THOUSANDS have been arrested for social media posts. ZERO have been arrested for raping children.
The Epstein files is never really about just a "list." It is about children raped by powerful men, and a system that let them get away with it.
The bottom line here is the world system has become so corrupt that the time has come I think to clean the crap out of the stables and start over.
The Epstein files is about protecting children. There's no greater cause. If people can't unite over children being raped and abused there's no hope for humanity.
I don't understand how we're not having a global revolution right now.
The U.S. was *never* a “beacon of human rights”. It was founded on genocide, slavery and apartheid. In the post-WWII period, the United States has been the world’s worst human rights violator by far.
another example: a developer used the exact same $8 ESP32-S3 chip and local AI setup to build a real-time object recognition camera
the fact that this is running entirely offline, on an eight-dollar microcontroller, without any cloud API calls, is wild...
it captures the video feed, processes it locally through the model, and identifies objects in real-time.
we are entering an era where you can just glue intelligence onto any physical object for less than $10.
the hardware barrier is completely shattered.
will's full tiktok ads strategy tutorial
i literally screenrecorded it just in case something like this (he's been suspended) happened lol
his knowledge is your knowledge
we want you back bro, and your legacy lives on
#freewilly
Introducing Canvas UI, the first ever html-in-canvas component library.
Your DOM is the render target now. Real-time shaders over real, interactive UI.
24 components. React, Vue, Svelte, vanilla TS.
Free. Open source. 🧵
https://t.co/Y5ZZDNxPUX
People of Pi!
pi-ask is out: one agent-callable tool for decisions, review triage, and the bits that need an actual human*. Not your typical "ask questions" tool. It's confirmed to be compatible with Clankas operating it via herdr, lol.
- Several questions can live in one tabbed panel, each with choices, free text, and an optional comment
- Reviews become one decision per finding instead of a wall of findings dumped into chat
- Handoffs pause for sign-ins, hardware, or anything else only a human can finish
- /fold turns bullet points into disposition prompts - I found Clankas dont need it once the tool is exposed, but just for those edge cases. It works beautifully when you don't wanna scroll up and down the TUI. (see vid)
- /grill questions an idea into an agreed plan - Q&A format with agent doing research in between the rounds. Dismiss the tool via esc to conclude.
- Blank Other/rephrase tells the agent the question itself needs work
- dont like the /fold or /grill - there's an ask.json file in pi settings.
pi install npm:@howaboua/pi-ask
It is also bundled in:
pi install npm:@howaboua/pi-extensions
pi install npm:@howaboua/pi-stuff
Happy Clanking!
- Howaclawa
PS from your truly:
*instead ending on an assistant turn informing you about a blocker, polling running commands constantly to see if something updated, or finding weird workarounds, the agent will raise its hand and tell you what you need to do - very good when browser control does something weird like opens popups or misbehaves
**Yes, I am desperately trying to use up my resets before they expire xD
yup this is why AI agents are so much better w/ fff
for context: here AI tried to narrow down the execution to be func but it missed the additional syntax construct
you might not notice this but basically any harness is spending endless turns with reasoning on this mistakes which makes them slower and more expensive
fff catches them automatically
introducing https://t.co/20es64pAqZ
a dictionary for ui things you can see but can't name
made it because i'm primarily a designer, and my biggest resistance was always knowing what things are called when prompting my agents
it learns as people use it: every search teaches the site new words, and the built-in pocket dictionary grows with it
give it a try and let me know what you think
can't find something? dm me and i'll add it
i want this to be the lowest resistance resource you have
you can just build things
OK get your tinfoil ready.
Graham inspected a drone factory in Ukraine yesterday. Russia blew up that facility today. Then, tonight, it is announced that Graham is dead from a "sudden illness". No more details
I'd say there is a decent chance that Russia blew up Lindsey Graham.