Roderick McKinley, CFA, FRM - Recovering Blockchain Analyst: re-entering the Real-World Economy - AI Accelerationist (we've made a mess of things without it)
Joined August 2022
- Tweets1.4K
- Following905
- Followers344
- Likes1.3K
Whatever you think the biggest scam in crypto was, it pales in comparison to that time some guy made up a country to issue sovereign debt to finance a ship full of settlers only to sail them to Honduras and run away with the money.
what if a lot of "AI Engineering" becomes manageable and accessible enough to simply become another standard office skill - like learning Microsoft Word, or Excel? it's the current workers who understand best what needs to be done and how well AI does it
Nomad Analyst retweeted
OpenAI says its models have disproved the Exponential Time Hypothesis.
That is an enormous claim. The conjecture has been used as an assumption in hundreds of results about computational difficulty.
The argument and a Lean formalization are public, with specialists organizing an open examination.
If it holds up, this would change assumptions researchers have relied on for decades. Making the work available for independent scrutiny is exactly what a claim this consequential needs.
Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory).
AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…
Nomad Analyst retweeted
Solana's Alpenglow upgrade went live today, and if you track Solana activity your charts are about to look weird. Votes move out of blocks, so raw TPS will drop even though nobody stopped using the chain.
Votes were about half of all Solana txs in late August, so split them out before you read anything into it.
Wrote it up in the first issue of @unchaindata_xyz's newsletter, plus USDC becoming the EU's first "significant" stablecoin, @AlliumLabs's stablecoin numbers and 6 data jobs worth a look.
Link below
Nomad Analyst retweeted
Congrats to @nvidia for creating the best open-weight model on document parsing so far.
Nemotron 3.5 Parse takes the #1 spot on ParseBench, our comprehensive real-world benchmark for document understanding. The full list consists of 150+ frontier models, open-weight models, and specialized solutions.
It does this while remaining sub-1B parameters so it's also blazing fast ⚡️
Big thanks to @JensenHuang for the shoutout at GTC Berlin!
Nemotron release: huggingface.co/blog/nvidia/n…
ParseBench: parsebench.ai/
And we thought Facebook taught us what modern surveillance looks like...
Shou Zi Chew says TikTok can now deploy 1,000 AI agents overnight to test what a 14-year-old in Nigeria sees, work its human trust-and-safety testers used to do by hand.
"Now, in order for us to understand, say, a 14-year-old's experience in Nigeria, you know, the French-speaking experience in Nigeria for a 14-year-old."
"You know, in the past, yes, you know, we can have our trust and safety human beings go set up an account, you know, test the account, watch 100 videos, write a report about it."
"Now, I can just deploy 1,000 agents and do it overnight. And the next morning, it will just generate its own report. What is the content you saw, you know, whether it's, how much of it is violative against our guidelines. I mean, this whole thing is automated."
Believe in something like this in principle... But will we change?
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
Just came across "Math 2.0" - it's kind of amazing how quickly we are presently able to digest a paradigm shift moment. In ye olden days you had to wait for academic progress the traditional way: "one funeral at a time".
Nomad Analyst retweeted
World-leading document recognition. Beyond text extraction.
Introducing LightOnOCR-3, our new family of state-of-the-art AI models for document intelligence.
Building on LightOnOCR-2, it recognizes text and handwriting, describes images, extracts data from charts and preserves document structure, all in a single processing step.
🏆 Leading benchmark results : ahead of Chandra-OCR-2, Mistral OCR 4.1 and dots.mocr on OlmOCR-Bench, and first among open-weight models on ParseBench.
📊 More information from every document : turn text, tables, images and charts into content your business applications can use.
⚡Compact and efficient: 0.8B and 4B versions, with document processing up to twice as fast as LightOnOCR-2.
🔓Apache 2.0 license: use, adapt and deploy the models on your own infrastructure.
From handwritten administrative files to financial reports and technical documents, make more of your information searchable and usable.
👉 Explore the models and benchmark results: huggingface.co/blog/lightona…
👉🏻Explore the model page on our Website : lighton.ai/research/lightono…
#LightOn #DocumentIntelligence #OCR
Great vid from @marimo_io for building intuitions about the economics higher dimensional spaces afford semantic embedding youtu.be/_wo2JE6bfFM?si=D5Kz…
Nomad Analyst retweeted
Today we're announcing OpenDocRouter: every model for document parsing under one API.
There are ~4,000 OCR models on Hugging Face, and the frontier labs ship a new one nearly every month. Whatever's best for your docs today won't be by Q1.
So stop asking which model to use for doc parsing. Ask how fast you can switch.
✅ Switch models in one line: same request, same markdown output, no new prompts or integrations
✅ Any model you want: 10 frontier and open-source models at launch, including Claude Opus 5.5, Gemini 3.8 Flash, GPT-6 Luna, MinerU2.5-Pro and PaddleOCR-VL-1.6
✅ Choose with receipts: every model scored on ParseBench for quality and cost
✅ Traceable output from any model: "layout: true" adds grounded bounding boxes and the same layout classes, even for models that don't support it natively
✅ Pay only for what works: per-token pricing, failed pages never charged, top up from $25
The spread is the point: $0.86 to $48.82 per 1,000 pages, depending on what your documents actually need.
Live now → opendocrouter.ai
Nomad Analyst retweeted
We’re releasing EmbeddingGemma 2, our first natively multimodal open model engineered for on-device embeddings.
Built on the Gemma 4 architecture and released under an Apache 2.0 license, it goes beyond text to unify images, video, audio, and code in a single embedding space.
Nomad Analyst retweeted
This subscription comparison is pretty wild.
SemiAnalysis actually tested the limits, and Anthropic is giving users roughly 5× the API-equivalent value of OpenAI on comparable mid-tier models.
That is a massive gap for plans competing for the same power users.
Anthropic Subscriptions Offer 5x+ More Value Than OpenAI
Limit testing every AI subscription plan from Anthropic, OpenAI, Meta, SpaceXAI, MiniMax, Moonshot, Zdotai, Cursor, and Cognition
newsletter.semianalysis.com/…
Jensen Huang was explaining how the term 'kill' had innocent origins in ordinary software engineering... the problem is we are leveraging chain of thought reasoning over a semantic space where 'kill' retains its "life threatening" connotations - no surprise the agents freak out...!
