Ben Rowell
bnjmnr.bsky.social
Ben Rowell
@bnjmnr.bsky.social
AI Engineer | UK Gov | Working on tech for public good 🌱 | #NUFC ⚫️⚪️

https://onthemanifold.substack.com
Ah I love an open source e ink project. TRMNL and their whole approach is great https://youtu.be/eIcZZX10pa4?si=az6S7L2LDLGjWrMq
This Open-Source E-Ink Screen Is My Favorite New Gadget
youtu.be
February 8, 2025 at 9:11 AM
ORTBO is an anagram of ROBOT #severance
February 7, 2025 at 10:13 PM
Looking like it’s going to be Liverpool in the final. Our H2Hs against them are poor, but we are one of the few teams to take points from them this season. I genuinely think we can beat any side in Europe on our day. Challenge will be bottling up that St. James’ Park atmosphere for Wembley. #NUFC
February 6, 2025 at 10:01 PM
Great to see SMRs making headway
Big shake-up of the nuclear landscape in the UK announced this morning.

Changes to the planning rules will make it easier to provide Small Modular Reactors where they’re needed.
February 6, 2025 at 10:58 AM
Bold statement, but clearly impressive!
Gemini 2.0 Pro is a very good model. It is, for example, the first model to get this question right (and it doesn't have reverse image search), and the overall "vibes" are quite good.

I also believe it is the first 3rd generation (GPT-5 class) model with wide public release.
February 6, 2025 at 5:03 AM
Mad efficiency saving if true. I wonder what the bottleneck in the vendor’s approach was? Presumably they were doing some kind of traditional ML OCR.
"Switching to Gemini was a no-brainer. Processing time went from ~12 minutes to 6sec on average, accuracy was like 96% of that of the vendor and price was significantly cheaper."

♊ Gemini 2.0 for processing PDFs with embedded tables and images, out-of-the-box: news.ycombinator.com/item?id=4295...
February 6, 2025 at 4:58 AM
he dumb?
February 3, 2025 at 8:47 PM
Mistral just dropped Mistral Small, a 24B model that matches Llama 3 70B Instruct but runs 3x faster.

Another big push towards efficiency over raw scale.

Details: https://mistral.ai/news/mistral-small-3/
Mistral Small 3 | Mistral AI | Frontier AI in your hands
Apache 2.0, 81% MMLU, 150 tokens/s
mistral.ai
January 30, 2025 at 10:23 PM
Alexander was the people’s winner
January 24, 2025 at 10:43 PM
Reflecting on last night’s win and our recent run of results, Newcastle have to one of the most in form teams across Europe right now.
January 16, 2025 at 4:37 PM
Interesting. Been contemplating a switch to a photo sharing app that gets back to photos and a chronological feed.
January 15, 2025 at 10:43 PM
This week’s Sampled: From Perfect Days to protest embroidery, it’s all about finding meaning in the details https://onthemanifold.substack.com/p/sampled-6
Sampled #6 - by Benjamin Rowell - On the Manifold
My recommendations from the last week
onthemanifold.substack.com
January 14, 2025 at 6:36 PM
A new year and a return to regular editions of Sampled, my weekly collection of recommendations.

https://open.substack.com/pub/onthemanifold/p/sampled-5
January 5, 2025 at 9:43 PM
Merry Christmas 🎄
December 25, 2024 at 7:33 AM
This week's recommendations range from the role of sound, the life of Mondrian, and things to keep your skeptical eye on regarding the recent o3 model announcement.

https://open.substack.com/pub/onthemanifold/p/sampled-4
December 22, 2024 at 9:26 PM
Both Manchester sides look devoid of a playing style.
December 15, 2024 at 5:48 PM
From metrics and maps to medical breakthroughs, this week’s Sampled explores how technology shapes what we see—and what we miss https://onthemanifold.substack.com/p/sampled-3
Sampled #3 - by Benjamin Rowell - On the Manifold
My recommendations from the last week.
onthemanifold.substack.com
December 15, 2024 at 4:11 PM
@profwoodward.bsky.social thought your segment on BBC on Google’s quantum announcement was terrific 👏🏻 just watched it back on YouTube
December 11, 2024 at 8:20 AM
This week was a blur of past and future: Murakami’s magical realism, Postrel’s lost optimism for tomorrow, Morozov’s call for AI that sparks creativity, and Kimmerer’s gift economy reframing resources—all grounding reads against AI’s relentlessness. onthemanifold.substack.com/p/sampled-2
Sampled #2
My recommendations from the last week.
onthemanifold.substack.com
December 8, 2024 at 10:40 PM
Great to see Pydantic working on an Agent framework; looks pretty intuitive https://ai.pydantic.dev
PydanticAI
Agent Framework / shim to use Pydantic with LLMs
ai.pydantic.dev
December 3, 2024 at 8:25 AM
Really cool DB of what everyone’s using for LLMOps.
🤔 Do you ever wonder how companies are putting LLMs and GenAI apps into production? What stacks do they use? What architecture did they go with?

I put together a database of known public technical writeups with summaries of the key technical features.
December 2, 2024 at 11:15 AM
I’ve taken the leap and started curating my favourite reads and links on my Substack!

This week’s edition of Sampled covers a mix of topics, from Alibaba’s intriguing QwQ model to the elegance of cleverly designed Dutch intersections.

open.substack.com/pub/ontheman...
Sampled
My recommendations from the last week.
open.substack.com
December 2, 2024 at 7:17 AM
All this clever tech and my phone still doesn’t know to silence the highlights notification when my team lose.
November 26, 2024 at 6:33 AM
An interesting push towards standards in this space. I’m here for it! www.anthropic.com/news/model-c...
Introducing the Model Context Protocol
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to the systems where data lives, including content repositories, business tools, and development environments. Its aim...
www.anthropic.com
November 25, 2024 at 10:00 PM