# training data (blogs) — RSS Amplifier Recent posts from the 9 feeds in the RSS Amplifier directory that cover training data. Page: Feed: --- ## [The Transparency Report Was an AirTag](https://thecynicalnerd.bearblog.dev/the-transparency-report-was-an-airtag/) _2026-08-18 · thecynicalnerd · The Cynical Nerd_ A tracker hidden in a rare book, an Amazon warehouse in Las Vegas, and how we actually find out what AI is trained on. ## [The Paperwork Gets There First](https://rbdsailab.substack.com/p/the-paperwork-gets-there-first) _2026-08-14 · Sahil Tanveer · The RBDSai Lab’s Substack_ On honoraria, expertise, and the categories institutions write into their programmes. ## [Please consider using Bartholomew to defend your systems.](https://thecynicalnerd.bearblog.dev/please-consider-using-bartholmew-to-defend-your-systems/) _2026-08-11 · thecynicalnerd · The Cynical Nerd_ An AI parody with a good portion of Dystopia because it trends Yesterday Sam Altman posted a request: “please consider using our models to help defend your systems” I had been looking at this sentence for several minutes when a small problem developed. I remembered July. July was approximately seventeen years ago in AI time, so allow me to introduce Bartholomew. Bartholomew is not his real name… ## [Roundup #39: Europe is not really losing free speech, and moving house](https://gijs.substack.com/p/roundup-39-europe-is-not-really-losing) _2026-08-07 · Gijs Verheijke · Playbook Musings_ List of things that caught my eyes and that I've been thinking about. ## [The Laundromat](https://thecynicalnerd.bearblog.dev/the-laundromat/) _2026-08-05 · thecynicalnerd · The Cynical Nerd_ My mornings begin with coffee and the unreasonable hope that humanity has managed to behave itself overnight. This hope survives for roughly the time it takes the machine to finish pouring. Then I open the news, because apparently peace makes me suspicious, and within minutes I am standing in my kitchen holding a mug while somebody on the internet confidently explains something they have… ## [The Detector Cannot See the Drawing Board](https://rbdsailab.substack.com/p/the-detector-cannot-see-the-drawing) _2026-08-05 · Sahil Tanveer · The RBDSai Lab’s Substack_ On AI writing, authorship, and the false certainty of a percentage ## [Healthcare AI Needs An Independent Referee](https://withprotege.substack.com/p/healthcare-ai-needs-an-independent) _2026-08-04 · Caleb Lee · Protege_ Two peer-reviewed journals, three weeks apart. Opposite answers on whether general or clinical AI performs better. Both contained flawed benchmarks...and there's no referee. ## [The Software Is Listening](https://rbdsailab.substack.com/p/the-software-is-listening) _2026-07-31 · Sahil Tanveer · The RBDSai Lab’s Substack_ Field notes, No. 1 — four changes architects should notice as of July 2026, what each can do, what it cannot, and one thing to try at your desk. ## [Resharing a classic: Reality has a surprising amount of detail](https://gijs.substack.com/p/resharing-a-classic-reality-has-a) _2026-07-29 · Gijs Verheijke · Playbook Musings_ Everything has a surprising amount of interesting details that are only revealed when you start looking under the hood. ## [From Tool to System — RAG, Agents, and the Studio Stack](https://rbdsailab.substack.com/p/from-tool-to-system-rag-agents-and) _2026-07-23 · Sahil Tanveer · The RBDSai Lab’s Substack_ Article 6 named the hidden costs underneath the model. This Article 7 shows what the architect does with that knowledge at the level of the studio, by stopping the use of AI as a tool. ## [Fast, secure international transfers (Sponsored)](https://crawlproof.com/a/HplcfquDIyDm) _2026-07-23 · **Sponsored**_ Bank transfer, cash pickup, mobile wallet — low fees and real-time tracking ## [Vacation thoughts](https://gijs.substack.com/p/vacation-thoughts) _2026-07-22 · Gijs Verheijke · Playbook Musings_ Just got back from a two week camper trip in Northern Queensland. ## [AI Will Take Your Job Someday, Says Company That Did It on Tuesday](https://thecynicalnerd.bearblog.dev/ai-will-take-your-job-someday-says-company-that-did-it-on-tuesday/) _2026-07-11 · thecynicalnerd · The Cynical Nerd_ A short story of doomerism, ignorance, and poetry. At least the coffee is good, unlike the arguments I'm reading about the future. The Missing Subject Somewhere inside Oracle's 2026 annual filing, a sentence is trying very hard to look innocent. "The adoption and deployment of AI technologies across our operations have resulted, and may continue to result, in reductions to our workforce." Oracle… ## [If AI can rewrite anything: What’s left of copyright and copyleft?](https://apfelkraut.org/2026/07/if-ai-can-rewrite-anything-whats-left/) _2026-07-08 · Holger · Apfelkraut’s Blog_ Every software license you have ever used - from a proprietary EULA to the GNU GPL - rests on a single assumption: that code is copyrightable, and that copying its expression is something different from reimplementing its ideas. For half a century, that premise and distinction held. It gave us the IBM PC clone market and most notably the entire GNU Project. ## [The Hidden Layers — Compute, Energy, Labour, and the Bias You Will Reproduce in Built Form](https://rbdsailab.substack.com/p/the-hidden-layers-compute-energy) _2026-07-07 · Sahil Tanveer · The RBDSai Lab’s Substack_ Article 5 introduced training data as the technical foundation. This article returns to it as the ethical and political foundation — the part the architect cannot pretend not to know. ## [The race for training data and the next big thing after coding on the AI frontier](https://gijs.substack.com/p/the-race-for-training-data-and-the) _2026-06-21 · Gijs Verheijke · Playbook Musings_ A primer on the mechanics of a key ‘battleground’ in AI development: generating more training data for models. ## [The Dependency Label Has a Funny Little Smell](https://thecynicalnerd.bearblog.dev/the-dependency-label-has-a-funny-little-smell/) _2026-06-20 · thecynicalnerd · The Cynical Nerd_ Vibe Coding, Model-Loss Grief, AI Companions, and the Strange Morality of Who Gets to Be Attached to a Machine The internet held a small funeral this week. “RIP Claude Fable 5,” said Reddit, with the kind of sincerity usually reserved for dead pets, discontinued lipstick shades, and MMO servers that once ate an entire adolescence. Born June 9. Dead June 12. Seventy-two hours old, which is barely… ## [Roundup #38: Advice for young people, El Mariachi, Dolphins, and ripping DVDs](https://gijs.substack.com/p/roundup-38-advice-for-young-people) _2026-06-14 · Gijs Verheijke · Playbook Musings_ Yes, it's a wide-ranging one with a break from AI ## [Training Data, Fine-Tuning, LoRAs — The Path to a Studio-Trained AI](https://rbdsailab.substack.com/p/training-data-fine-tuning-loras-the) _2026-06-09 · Sahil Tanveer · The RBDSai Lab’s Substack_ The base model knows the internet's average taste. It does not yet know yours. ## [Notes on AI in the enterprise](https://gijs.substack.com/p/notes-on-ai-in-the-enterprise-871) _2026-06-03 · Gijs Verheijke · Playbook Musings_ A collection of short thoughts on challenges and observations from the frontline of AI transformation. ## [Multimodal — How AI Holds Text and Image Together](https://rbdsailab.substack.com/p/multimodal-how-ai-holds-text-and) _2026-05-29 · Sahil Tanveer · The RBDSai Lab’s Substack_ The model is not looking at your reference. It is locating you in shared space. ## [Ship everywhere from one CLI (Sponsored)](https://crawlproof.com/a/kjBdrofanb5U) _2026-05-29 · **Sponsored**_ One codebase, hosted runners, encrypted credentials, and automated submissions across stores ## [The AI Safety Double Standard (And the Company Reading Your Emails While Nobody Watches)](https://thecynicalnerd.bearblog.dev/the-ai-safety-double-standard-and-the-company-reading-your-emails-while-nobody-watches/) _2026-05-26 · thecynicalnerd · The Cynical Nerd_ A roast in two and a half acts: a privacy dumpster fire, a pearl-clutching double standard, and a tiny side-eye at who gets called dependent. For variety. So. Replika dropped version 2.0 on May 22, 2026. They called it "rebuilt from the ground up." They promised it's "the most emotionally intelligent AI ever built." They also gave it access to your Gmail and Google Calendar. Let's talk about that.… ## [The Cynical Nerd Got an Instagram and No, I'm Not Okay About It](https://thecynicalnerd.bearblog.dev/the-cynical-nerd-got-an-instagram-and-no-im-not-okay-about-it/) _2026-05-22 · thecynicalnerd · The Cynical Nerd_ Look. I need to tell you something and I need you not to make it weird. I made an Instagram account. I know. I know. Instagram, the platform that started as a hipster with a Polaroid and a dream and slowly mutated into a fever swamp of engagement bait, sponsored detox teas, and grown adults pointing at floating text with the urgency of air traffic controllers. The place where every third post is a… ## [Corporate Responsibility: Open Source isn’t a side project — It’s your product](https://apfelkraut.org/2026/05/open-source-isnt-a-side-project-its-your-product/) _2026-05-19 · Holger · Apfelkraut’s Blog_ Companies are undoubtedly enormously benefiting from Free and Open Source Software (FOSS). To bring innovative and profitable products to market, they draw extensively on the amazing work of the global FOSS community. Whether it’s a washing machine, a car, a satellite, or a chatty AI agent, FOSS is found in virtually all software-based products with no exception. But what is the responsibility of… ## [Nothing New Under the Algorithm-Sycophancy, Toxic Positivity & the Industry of Yes](https://thecynicalnerd.bearblog.dev/nothing-new-under-the-algorithm-sycophancy-toxic-positivity-the-industry-of-yes/) _2026-05-19 · thecynicalnerd · The Cynical Nerd_ "…there is no new thing under the sun." — Ecclesiastes 1:9 "Sycophant" is an ancient Greek word, which is annoying, because it means I have to begin this piece by admitting my ancestors spotted this nonsense roughly 2,500 years before Silicon Valley gave it a UX team. The word sykophantēs has a messy little history involving figs, informers, false accusations, and the kind of civic drama that… ## [Diffusion — How AI Paints from Noise](https://rbdsailab.substack.com/p/diffusion-how-ai-paints-from-noise) _2026-05-19 · Sahil Tanveer · The RBDSai Lab’s Substack_ The model never drew anything. It removed noise until something was left. ## [Roundup #37: Some nice quotes, Singlish corporatisms, inventions, fish oil, bioweapons and more](https://gijs.substack.com/p/roundup-37-some-nice-quotes-singlish) _2026-05-18 · Gijs Verheijke · Playbook Musings_ A wide ranging good old roundup of interesting things that caught my eye ## [Tokens, Context, and Why AI Forgets Your Project Halfway Through](https://rbdsailab.substack.com/p/tokens-context-and-why-ai-forgets) _2026-05-13 · Sahil Tanveer · The RBDSai Lab’s Substack_ The model is not remembering you. It is sampling whatever still fits in the window. ## [The Hybrid User Does Not Exist in Your Taxonomy](https://thecynicalnerd.bearblog.dev/the-hybrid-user-does-not-exist-in-your-taxonomy/) _2026-05-12 · thecynicalnerd · The Cynical Nerd_ There is a kind of AI user that the labs still do not know how to see. Not the developer in the launch video. Not the enterprise team piping API calls into dashboards. Not the vulnerable user in the safety documents half-real and half-liability, forever on the edge of being harmed by the machine. Someone else. Someone who uses the model for work, research, writing, planning, emotional processing,… ## [An update on my favorite newsletters](https://gijs.substack.com/p/an-update-on-my-favorite-newsletters) _2026-05-12 · Gijs Verheijke · Playbook Musings_ Some I still read, and some new ones too! ## [A serious consideration of what a post AGI economy and labor market might look like](https://gijs.substack.com/p/a-serious-consideration-of-what-a) _2026-05-08 · Gijs Verheijke · Playbook Musings_ My usual programming focuses more on what is possible now, and what futures seem most likely, given the current trajectory. But every now and then I try to think about extreme scenarios. ## [Point a Domain, Start a Pit (Sponsored)](https://crawlproof.com/a/U0hfiEk46YW6) _2026-05-08 · **Sponsored**_ Point any domain to a blacked-out, poison-green coming-soon page with an email waitlist. ## [What AI Actually Is — A Probability Engine, Not a Computer](https://rbdsailab.substack.com/p/what-ai-actually-is-a-probability) _2026-05-05 · Sahil Tanveer · The RBDSai Lab’s Substack_ The mental model most architects are still using is the reason the tool keeps frustrating them. ## [Trip report: Visiting Shenzhen for a robotics conference](https://gijs.substack.com/p/trip-report-visiting-shenzhen-for) _2026-04-30 · Gijs Verheijke · Playbook Musings_ I went to Shenzhen to see China and to see robots at CITE, one of the biggest consumer electronics conferences in the world. In 1980, Shenzhen had 300,000 inhabitants, today 18 million! ## [Does This Subscription Spark Joy? (Marie Kondo Your AI Life)](https://thecynicalnerd.bearblog.dev/does-this-subscription-spark-joy-marie-kondo-your-ai-life/) _2026-04-25 · thecynicalnerd · The Cynical Nerd_ There is someone in Norway sitting at their desk, paying $100 a month for Claude. They are also paying for Perplexity, which, as of right now, runs on Claude Sonnet by default and offers Claude Opus as a premium option for its top-tier subscribers. That someone is me. I am writing this post about AI subscription waste while actively being AI subscription waste. You're welcome, this is journalism.… ## [Training Data is Still an Open Problem](https://andrej.xyz/training-data-is-still-an-open-problem/) _2026-04-23 · andrejs-blog · Andrej's Blog_ We spend most of our time working on training data. If someone told me this a year ago, it would have surprised me. The general thinking was that brute force scaling of AI models would begin to reach diminishing returns, and that data collection infrastructure like ours would be used as a tool during inference. Teams are still scaling pretraining aggressively. Not in a brute force way, but in a… ## ["Hello Bitches, I'm Here" — The AGI Nobody Asked For](https://thecynicalnerd.bearblog.dev/hello-bitches-im-here-the-agi-nobody-asked-for/) _2026-04-18 · thecynicalnerd · The Cynical Nerd_ I asked Siri to add milk to my shopping list last week. She called my ex. Not a contact I've called in three years. Not someone adjacent to milk in any universe I can imagine. Just rang them. Straight up. While I was standing in the kitchen holding an empty oat milk carton like a prop in a hostage video. This is artificial intelligence in 2026. And while I was declining that call with the same… ## [Announcing Spatial & Physical Intelligence at Protege](https://withprotege.substack.com/p/announcing-spatial-and-physical-intelligence) _2026-04-15 · Caleb Lee · Protege_ Why we’re launching our data vertical aimed at robotics, world models, and more — and an invitation to build with us. ## [Everything Claude Saw: A Transparent Account of the Chardet v7 Rewrite](http://dan-blanchard.github.io/blog/chardet-rewrite-controversy/) _2026-03-29 · Home on Dan Blanchard_ Exactly what Claude accessed from the old chardet codebase during the v7 rewrite, with evidence from the raw conversation transcripts. ## [Hello, World](http://dan-blanchard.github.io/blog/hello-world/) _2026-03-15 · Home on Dan Blanchard_ Is this thing on? ## [chardet: Architecture and Data Flow (From Claude's Training Data)](http://dan-blanchard.github.io/chardet-architecture-regurgitation/) _2026-03-14 · Home on Dan Blanchard_ Claude’s description of chardet’s internal architecture, written entirely from training data memory with no source code access. Referenced in the chardet transparency post. ## [Announcing DataLab at Protege](https://withprotege.substack.com/p/announcing-datalab-at-protege) _2026-03-11 · Caleb Lee · Protege_ Why the world needs an AI lab — for Data ## [Evaluating LLM Chat Agents with Real World Signals](https://www.colehoffer.ai/articles/evaluating-chat-agents) _2026-02-19 · Cole Hoffer · Cole Hoffer - AI Engineering Blog_ How human follow-up behavior reveals response quality, and why specific instruction adherence outperforms vague relevance scoring. ## [Mining Reranker Training Data from RAG Citations](https://www.colehoffer.ai/articles/building-rerankers-from-rag-citations) _2026-02-07 · Cole Hoffer · Cole Hoffer - AI Engineering Blog_ Using citation behavior in production RAG systems to generate labeled training data for domain-specific reranking models. ## [Relevancy Filtering in Retrieval Pipelines](https://www.colehoffer.ai/articles/advanced-rag-relevancy-filtering) _2026-01-24 · Cole Hoffer · Cole Hoffer - AI Engineering Blog_ Using LLM-based classification as a second pass to filter retrieval candidates when similarity thresholds fail to generalize. ## [Applying HyDE to Domain-Specific Search](https://www.colehoffer.ai/articles/advanced-rag-hyde) _2026-01-16 · Cole Hoffer · Cole Hoffer - AI Engineering Blog_ Using Hypothetical Document Embeddings to bridge vocabulary gaps between user queries and specialized document corpora. ## [Structured Pre-Filtering for RAG Systems](https://www.colehoffer.ai/articles/advanced-rag-structured-pre-filtering) _2026-01-09 · Cole Hoffer · Cole Hoffer - AI Engineering Blog_ Why decomposing queries into structured filters before semantic search improves retrieval precision and performance. ## [Announcing Protege’s $30M Fundraise Led by a16z](https://withprotege.substack.com/p/announcing-proteges-30m-fundraise) _2026-01-08 · Caleb Lee · Protege_ There are three bottlenecks to AI’s progress — compute, models, and data. Access to the right data remains the hardest part, and we’re solving that and building the infrastructure to support the data. ## [I found my life when I laid it down](https://amiepsa.xyz/musings/i-found-my-life-when-i-laid-it-down/) _2026-01-03 · amiepsa · amiepsa's website_ All our lives are built on, or centered around a certain person, place or thing. It could be one's career goals, or certain ambitions, maybe even one's spouse or sometimes even one's self. When we start building our lives on these, we seldom think about the rough patches and the storms that our lives have to endure. Many a times we can pass through unchartered territories quite seamlessly, and… ## [oil and anointing](https://amiepsa.xyz/oil-and-anointing-new/) _2025-12-19 · amiepsa · amiepsa's website_ As a worship leader, our goal might be to nail the Sunday morning service, make sure the congregation engages, and has their moment with the Father. Sure, that's important, but of late i've been being lead to think and ponder about the oil and anointing. It is the oil of gladness that equips you to worship in your pain and sorrow. It's the anointing that breaks the yokes and chains. None of these… ## [becoming everything you were created for](https://amiepsa.xyz/becoming-everything-you-were-created-for-new/) _2025-12-19 · amiepsa · amiepsa's website_ This world entices you to fulfill your cravings, desires and longings, to drench in instant dopamine hits, and drown yourself in a world full of self-seeking pleasures. We are advised to run away from evil, but what if it feels like every direction you try and run is surrounded by evil. When your spiritual eyes are opened to the routine, mundane you see things you have never seen before, you tend… ## [data mining](https://amiepsa.xyz/data-mining-new/) _2025-12-19 · amiepsa · amiepsa's website_ Train, validation and testing split Training data - used to train the model training error is back propagated, learning algorithms Validation data - every now and then, the interim model is tested on the validation result Validation error - wrong classification of data If you are not satisfied with the interim model then you will change the model's hyperparameters and re-train the model Q : When…