Sign in

Olatomiwa Bifarin

@bifarinthefifth.bsky.social
1 followers 3 following 143 posts

essayist @ youtube.com/@epistemeEngine AI blog @ theepsilon.substack.com μgrants @ mbbgrant.substack.com

PostsRepliesMedia
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 23/08/2025
A couple days ago, I attended the IBM Tech Exchange Dev Day at Georgia Tech on Open Source AI. I ended up learning more than I thought I would. From Docling to Generative Computing, my notes: theepsilon.substack.com/p/from-gran...
theepsilon.substack.com
From Granite to Generative Computing
IBM’s Foundational Models, Unstructured Document Processing, and Generative Computing.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 16/08/2025
The latest pulse, Issue 28: GPT-5👨🏿‍💻, Comet🌐, and How to Publish for Machines📄 theepsilon.substack.com/p/pulse-iss...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 19/07/2025
It's already clear that knowledge works are going to undergo a massive restructuring (to say the least) over the next decade; many (most notably in coding) are already well under way. And knowledge work does not preclude running a lab & doing science.
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 12/07/2025
My latest AI reading list ε Pulse: Issue 27 -> Context Engineering👷🏿, Grok 4📒, and a Foundation Model for Tumor Microenvironment 🎗 theepsilon.substack.com/p/pulse-iss...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 10/07/2025
What happens when an AI runs a business? Anthropic put Claude in charge of a real-world vending machine for their "Project Vend" experiment.
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 07/07/2025
Folks at Cognition argues against building multi-agent AI systems. The author argues that decomposing tasks for multiple agents creates more problems (like context sharing) than it solves. More here: cognition.ai/blog/dont-b...
020
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 05/07/2025
Anthropic engineers' take on building multi-agent research system to tackle complex, open-ended questions. A lead agent delegates tasks to sub-agents who research in parallel in an orchestrator-worker agentic pattern
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 04/07/2025
Fine essay here on Bio-ML and Noetik (a startup building foundation models of tumor microenvironment) The thesis: The future of bio-ML is using rich, multimodal data to build foundation models for complex diseases, rather than focusing on simpler problems. www.owlposting.com/p/joining-n...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 03/07/2025
My latest academic work on applied NLP is now published in Analytical Chemistry: "A Large Language Model–Powered Map of Metabolomics Research" Paper: pubs.acs.org/doi/10.1021... App: metascape.streamlit.app/
011
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 28/06/2025
my notes on topic coherence measures in NLP, over at my AI blog. theepsilon.substack.com/p/topic-coh...
theepsilon.substack.com
Topic Coherence Metrics for Dummies
A non-technical treatment of topic model evaluation
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 27/06/2025
Andrej's analogy game is a hard. I really enjoyed listening to his talk on software in the age of AI. Highly recommend, if you haven't watched it already. www.youtube.com/watch?v=LCE...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 26/06/2025
We traded calloused hands for blue-light screens, thinking knowledge work would end human suffering. It eased plenty no doubt; yet we swapped worn-out bodies for sedentary-induced metabolic ‘fallouts’ (obesity, diabetes, and the like.), and all the suffering downstream. Progress with a hidden bill.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 24/06/2025
I came across this nice talk the other day: www.youtube.com/watch?v=uFP... Building effective agents with mcp via an agent library "mcp-agent"
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 24/06/2025
instagram.com/epistemeengine
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 21/06/2025
This is a good (video) podcast on Mech Interp, highlighting recent publications from Anthropic: www.youtube.com/watch?v=9YQ... Their research aims to move beyond treating LLMs as black boxes and instead reveal their "intermediate thought processes" via attribution graphs.
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 21/06/2025
What I am reading this summer: www.epistemeengine.com/p/what-i-am...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 20/06/2025
Sam Altman on the "Gentle Singularity": blog.samaltman.com/the-gentle-...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 20/06/2025
Which startup are your investing in? Cogito, The Prince, Form Room, Proletariat Perks, or Sisyphus
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 18/06/2025
The other day, I used a prompt to write a prompt that then writes another prompt. I suppose you can just do stuff these days.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 18/06/2025
instagram.com/epistemeengine [1]: www.nature.com/articles/s4... [2]: jamanetwork.com/journals/ja...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 16/06/2025
It takes some time for it to sink in but properly viewed, productivity isn’t about doing more; it’s about ignoring nearly everything, and that precisely, is what makes it incredibly hard.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 11/06/2025
instagram.com/epistemeengine
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 10/06/2025
instagram.com/epistemeengine
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 07/06/2025
🧵👇🏿
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 06/06/2025
Powerful action space, decentralization, privacy & security. A clear-eyed argument for personal, local, private AI agents by soumith chintala www.youtube.com/watch?v=jMo...
youtube.com
Personal, Local, Private AI Agents: Soumith Chintala
AI Agents are being worked upon by lots of people. But can they run locally, fast, keep your information access in your control and help you trust them with ...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 06/06/2025
Interesting, and unsurprising findings here on genAI usage at work: www.axios.com/2025/05/29/... The speed of change is so much that, I wonder how many "leaders" are able to keep up, talk less of writing useful guidelines. Keeping up is *literally* a day job.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 06/06/2025
instagram.com/epistemeengine [1]: www.worldobesity.org/about/about... [2]: www.who.int/news-room/f...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 05/06/2025
The next wave of commerce would be agentic. From malls to website to apps, and now agents, and this essay argues correctly for the case (with a more developed thesis): www.digitalnative.tech/p/agentic-c...
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 02/06/2025
We are at the verge of a full-blown prompt economy.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 31/05/2025
1/ I recently completed Alan Lightman's recent book, "The Miraculous from the Material." I enjoyed reading it but not without some qualms. I had to write a video essay + some philosophical commentaries, contrastingly titled, "The Material from the Miraculous" 🎥: www.youtube.com/watch?v=5hC... 🧵
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 29/05/2025
Boethius: A person is an individual substance of a rational nature. Made with Veo 3. This model will supercharge video essayists' workflows.
010
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 24/05/2025
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 22/05/2025
Apparently LangGraph agents can now be exposed as MCP tools. Exciting to see how the entire agent engineering ecosystem is rapidly maturing. doc: langchain-ai.github.io/langgraph/c... video: www.youtube.com/watch?v=AR4...
youtube.com
LangGraph Platform: MCP Support For Your LangGraph Agents
In this video, we'll show how LangGraph Platform integrates with MCP (Model Context Protocol). This enables every LangGraph agent to function as an MCP serve...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 18/05/2025
I first covered MCP when it was released in late 2024 and revisited at least once afterwards. Since it's gained popularity in the AI community, I thought it deserved a dedicated Pulse entry for a deeper dive. ε Pulse: Issue 25 → MCP Special + theepsilon.substack.com/p/pulse-iss...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 16/05/2025
1/ Finished up "AI 2027" recently It is my recommended reading for anyone with even the slightest interest in AI forecasting and safety. My summary & reflection: theepsilon.substack.com/p/the-intel... Some quotes from the scenario: 🧵 ai-2027.com/
120
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 15/05/2025
I can think of exceptions, but in general the intuition here is 🎯
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 12/05/2025
On AI writing detection: my intuition is that with increasing access to cheap, powerful AI (+ improved prompt eng. education): false negatives > false positives. (i.e., it will become easier to miss AI-generated writing than to misidentify genuine human writing.),
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 10/05/2025
An obvious hallmark of highly productive people today: avoiding *distractions & constant task-switching like a plague. My latest video essay explores this. www.youtube.com/watch?v=PiB... The essay is an excerpt from my book Wander: epistemeengine.gumroad.com/l/wander
epistemeengine.gumroad.com
Wander: Plans Are Overrated
About the Book📕In this unconventional memoir-essay hybrid, I write about a philosophy of how to approach knowledge work using my experience during my science graduate training as a springboard. Along the way, I confronted scientific dead ends ⚠️, grappled with career decisions that veered wildly off-script, and ultimately discovered the value of embracing uncertainty ✨. As I wrote, I ended up with a thesis statement: “Plans are Overrated.” This is not to say plans aren’t important; rather, they’re not as important as you think—if you think they’re terribly important. Through anecdotes and mini-essays, Wander explores: How to think about your “information diet” 🍽️ Practical productivity principles for modern knowledge workers. A philosophy of approaching failure that keeps moving you forward. The value of curating optionalities in an ever-changing world 🌎 Wander is equal parts reflective memoir and knowledge-work manifesto – a 10,000-word essay, short enough to be read in an hour. Wheth
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 03/05/2025
Around the Web Issue 36: The Next Pope. 🦠Recipe for a Cell. 🔰 AI Assistant for Avid Readers. 💻 A.I. Is Changing How Folks Builds Start-Ups. 🌎Ten Charts That Show How the World is Changing www.epistemeengine.com/p/around-th...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 30/04/2025
I'm currently reading a fine book on LLM finetuning by Voigt Daniel. Sometimes, jumping from one tutorial to the next can be ineffective. I’ve found this book helpful for reviewing all the key aspects of the subject. I plan to share my notes on my blog once I’m done reading.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 29/04/2025
Anyone who has spent less than 30 mins probing Gemini 2.5 Pro or o3 would know that these models are not nothing, agentic at core and extremely capable (2.5 Pro is my favorite, by a small margin).
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 28/04/2025
1/ When I first tried out GPT-3 circa September 2022 (via the Playground), a few months before ChatGPT came out, it was clear to me that this is fundamentally a technology that's going to upend the state of affairs.
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 25/04/2025
AI systems are getting a massive upgrade, but it’s not making the splash it deserves—too much at once, and too many ghiligi-style memes :) Here’s a great podcast episode on AI-enabled biology and medicine, highlighting some of the latest advances. www.youtube.com/watch?v=Jo3...
youtube.com
New in Nature: Google Agents Beat Human Doctors, Make Scientific Discoveries – With Vivek and Anil
In this episode, Nathan Labenz speaks with Vivek Natarajan and Anil Palepu from Google DeepMind about their groundbreaking work on AMIE (Articulate Medical I...
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 18/04/2025
I wrote this app prototype to solve a personal habituated pain point.  To stop reading in a circle, find knowledge gaps, read on-demand, and dig in.  Missing Pages: Towards an AI agent for avid readers.  Back Story: www.youtube.com/watch?v=4wn... Waitlist: tally.so/r/wgOJvM
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 16/04/2025
Forget secrets hidden from others. In a sense, the biggest secrets are the ones we hide from ourselves. I say more in my latest (short) video essay: What Are You Hiding? www.youtube.com/watch?v=BeO...
youtube.com
What are you Hiding?
Follow me on Instagram: https://www.instagram.com/epistemeengineSubscribe to my substack: https://www.epistemeengine.com/
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 13/04/2025
I have now had sufficient play time with Gemini 2.5 Pro. It is too good. While I believe we will still have better models, these things are hard to imagine.
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 12/04/2025
My latest AI newsletter -> ε Pulse: Issue #24 AI Paper Finder📄, TxGemma💊, and the Future of SaaS 📱 theepsilon.substack.com/p/pulse-iss...
000
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 12/04/2025
Dario Amodei predicted a "compressed 21st century," where AI-driven biology accelerates 50-100 yrs of progress into just 5-10 yrs. I was skeptical when I first read this prediction, and I remain so, although less so than before, largely because of some recent advances.
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 11/04/2025
Described as the USB-C port for AI apps, MCP is the open standardized protocol that has gained lot of attention lately. It connects AI apps (IDE, Claude Desktop) 🔌 to various data sources, tools, etc. If you say MCP is like a magic wand for LLMs, you might just be right.
100
Olatomiwa Bifarin @bifarinthefifth.bsky.social · 09/04/2025
Some Computer Use Agents (LLM that can *use* mouse and keyboard like you and me.) dev tools I stumbled on recently. The space seems to be maturing pretty well. Amazon's Nova Act SDK: github.com/aws/nova-act LangGraph CUA: openai & scapybara: www.youtube.com/watch?v=ndC...
youtube.com
LangGraph Computer Use Agents
A LangGraph agent that uses the OpenAI Computer Use API and Scrapybara to build a computer use agentpip install langgraph-cuaCode: https://github.com/langcha...
000