Tootfinder

'Matlin and Chakraborty designed a study around influence functions, a technique that estimates how much any single training document – an article, post, or page a model trained on – shaped how the model answered a particular benchmark question. Add those estimates up across the data, and you can see which kinds of writing – literary, technical, and so on – support a given ability in the model. But that holds only if you can confirm the model actually trained on the documents in question…

Shopware 6: "Can the Store API read this internal field? Can the Admin API write it?" A plugin's answers were "yes" and "yes, and then it gets overwritten anyway." Two undocumented flags solve this declaratively, with zero subscribers — the core uses them for order totals.

Shopware 6 Hidden Gems #14: WriteProtected & ApiAware — field-level security without a single subscriber - Winkelwagen GmbH - an ecommerce agency
A code review scenario. A plugin adds a purchasePrice-style field to a custom entity — internal margin data. The review question: "Can the Store API read

Google plans to launch experimental satellite MVP on a SpaceX rocket on October 1, as part of Project Suncatcher; MVP has four TPUs and will operate for a year (New York Times)
https://www.nytimes.com/2026/09…

Google Is Sending an A.I. Data Center to Outer Space
Next Thursday, Google is sending an experimental satellite into orbit that will have enough computing power to answer simple A.I. queries from space.

My wife nerd sniped me into answering some deep questions about Criminal Minds using nothing but the subtitles.
It's the worst case scenario docs for my RAG system so tried to solve that using some Graph storage and LLMs to make sense of the subtitle files.
Now I can answer things like "list all the episodes with female unsubs" etc
Not yet wrote the code - but Claude did a POC in Python and it worked \o/
Here's a short video,

When you think big tech/ai heads won't make you more angry, you learn they go after rare books, scan them and destroy original copy...
And this is considered "fair use" of course 🤦
https://pivot-to-ai.com/2026/07/23/ai-vendors-ar…

AI vendors are destroying rare books to feed the chatbot
Every AI vendor trains their chatbots on the Common Crawl, a copy of the whole public internet. But that’s just the starting point. AI vendors are desperate for more data — in the hope of making th…

inploid: Inploid: an online social Q&A platform
Inploid is a social question & answer website in Turkish. Users can follow others and see their questions and answers on the main page. Each user is associated with a reputability score which is influenced by feedback of others about questions and answers of the user. Each user can also specify interest in topics. The data is crawled in June 2017 and consist of 39,749 nodes and 57,276 directed links between them. In addition, for …

Iran's Masoud Pezeshkian addresses UN General Assembly
Iran’s Masoud Pezeshkian is speaking now.
“The US president described us as terrorists. We have been the victims of terrorism,” he says.
He says the US “bombed a school” and “children perished under bombs with weapons used by the US and Israel.”
He says Iran “did not bow our head … and responded, but we did not hit civilian populations,” but “only defended ourselves – we are not terrorists.”
As he goes on, th…

US delegation walks out as Iran’s Pezeshkian angrily responds to Trump at UN General Assembly - live
Iranian leader says ‘US president described us as terrorists. We have been the victims of terrorism’

'Matlin and Chakraborty designed a study around influence functions, a technique that estimates how much any single training document – an article, post, or page a model trained on – shaped how the model answered a particular benchmark question. Add those estimates up across the data, and you can see which kinds of writing – literary, technical, and so on – support a given ability in the model. But that holds only if you can confirm the model actually trained on the documents in question…

How a Georgia Tech team used the open Olmo stack to trace social reasoning | Ai2
A Georgia Tech team used Ai2’s fully open Olmo stack to trace social reasoning back to the training data that shaped it, finding that dialogue-rich, interpersonal writing had an outsized influence on the capability.

Sources: AI training data startup Micro1 raised $100M at a $4B valuation, up from $500M in September 2025 (Anna Tong/Forbes)
http://www.forbes.com/sites/annatong/2026/09/22/this-25-year-old-raised-ove…

This 25-Year-Old Raised Over $100 Million For His AI Startup Micro1
At the beginning of last year, Ali Ansari’s company, Micro1, was a recruiting business generating $7 million in ARR. It’s now generating over $500 million in ARR.

Docs: AI data startup Mercor had $614M in gross revenue in H1 2026, up 70% from all of 2025, with ~91% from foundation model makers like OpenAI and Anthropic (The Information)
https://www.theinformation.com/articles/mercors-fast-growth-r…

Mercor’s Fast Growth Relies on Biggest AI Companies, Documents Show
Mercor, a three-year-old data startup whose army of contractors teaches AI to give better answers, is growing fast but relies heavily on revenue from AI foundation model companies, financial documents seen by The Information show. The company generated $614 million in gross revenue in the first ...