Should be a banner at the top. Or settings on bottom of control panel on left

Did you add an API key?

Automating idea generation and execution with AI is getting pretty good. Starting to feel like a productive supervisor - student research relationship. It will likely change science enormously 🧪

open.substack.com/pub/selfsupe...

Just cancelled ChatGPT Pro & Cursor Pro

Not getting $40/m value from them when I can use Gemini 3 Pro +Antigravity (with Opus 4.5)

I suspect many others will do the same. If Google can undercut price and offer the same product (arguably better), others need to exceed expectations to get me back

Yes, and many bad uses of computers too.

Elegance and Unification in AI What fundamental physics taught me about where look for weaknesses

I started my substack this time last year when I had a few days off to think. I’ve had less to than I’d like in 2025 and that looks set to continue. I have a few ideas brewing but in the interim why not take a look at the post that kicked it off if you haven’t already 🧪

substack.com/home/post/p-...

New blog post about a side project I've been tipping away at.

I've build an app to explore open-ended idea discovery using LLMs in a multi-agent evolutionary algorithm.

I've also open source the code so you can play with yourself

Codex in particular is very slick at inviting you to farm out lots of small tasks in parallel.

It's all very cool, but so far I'd only be comfortable with using it for very tiny, easily verifiable changes [I merged ~6/7 PRs of tiny fixes after about 5 mins work from me].

They're both asynchronous coding agents that spin up a VM to make changes, test them, commit them and create PRs. They discuss their changes, and show you nice diffs.

This means you can just use your phone if you like and code from ... anywhere.

Might be a lack of imagination on my part but my default expectation is for the $6.5bn acquisition on Jony Ive's io to result in OpenAI wearable pendant like the so many already out there (e.g. Limitless below)

Hopefully the was something more exciting in the deal.

AlphaEvolve vs LLM skeptics Prompting our way to new discoveries

open.substack.com/pub/selfsupe...

It's an exciting time for scientists everywhere 🧪

I find this kind of science very exciting because it's accessible to those who can't build the models themselves.

deepmind.google/discover/blo...

I cringe when I see headlines from OpenAI leaders about "colonizing the light cone". They will probably continue to get very rich from monthly subscriptions. But it's the science led orgs like Deepmind that will change the world.

But Deepmind have been steadily closing the gap on the models front and are just in a league of their own on extracting real tangible value in Science and Engineering.

Yes, OpenAI are better at product and leaped ahead with their LLMs. I still prefer using ChatGPT to the alternatives.

Work like AlphaEvolve is what for me has always been the clear distinction between Google DeepMind and OpenAI. 🧵

Fantastic. Thanks so much for sharing

Ok, thanks, recordings would definitely be appreciated. We're working on FM for Neonatal EEG so I'm interested to hear about other's experiences

Looks really cool. Any remote option?

Cork medtech Cergenx secures FDA support for neonatal brain injury device UCC spin-out’s Wave device uses AI to offer hope of early intervention for newborns with brain injury

Delighted to see our Wave device receive breakthrough designation from the FDA

#neosky #medsky

www.irishtimes.com/business/202...

This represents a significant milestone for the field with the first ever demonstration of human expert level performance on held out datasets.

First of many for CergenX with our friends in INFANT Research Centre.

Thanks to our co-authors Geraldine Boylan, Sean Griffin, Aurel Luca, Sean Mathieson, and Soraia Ventura.

Supported by Enterprise Ireland DTIF grant

CergenX Blog - Scaling AI for expert level seizure detection in newborns Innovation in Neonatal Brain Assessment Technology.

Our seizure detection paper is finally out! Published in npj Digital Medicine.

Lots of work went into this one co-led with @johnotoole.bsky.social.

Blog: www.cergenx.com/blog/scaling...
Paper: www.nature.com/articles/s41...
#Neosky

What Lies Beneath: How Prompts Shape and Limit LLM Potential Our current models are more capable than we know.

😅 definitely part of it. But I think of it more a breakthrough in the crappy post training for prompting issue LLMs have. I wrote a bit more about it here if you're interested open.substack.com/pub/selfsupe...

Screenshot of open roles at Fauna Robotics

Want a job in robotics in New York? faunarobotics.com

"Our robots are safe, lightweight, and inexpensive - perfect for deploying novel AI algorithms, testing new capabilities among people, and collecting data at scale"

Sounds amazing. Target price point? Would love a hobby robot to train at home.

Interesting paper showing how LLMs change their representational geometry in-context to match a task structure: arxiv.org/abs/2501.00070

Here, the model’s latent representations show a grid structure matching the task.

#MLSKy #NeuroAI

It's an interesting idea to be able to ingest a someones feed and have an LLM judgement of whether they meet some niche category of interest or personality type. Basically automated personal.