Blaise Thomson

@b1rdmania @PrimaMente We’re here too! @bitfount - running federated AI infrastructure for clinical research

15 weird London AI companies you've probably never heard of London's AI scene isn't just chatbots and SaaS. Some genuinely strange stuff happening. @PrimaMente – building "foundation models for the brain" to study Alzheimer's and Parkinson's. Hologen – frontier medical AI spinout tied to UCL/King's. Eric Schmidt involved. hologen.ai @baseimmune – AI-designed vaccines targeting rapidly mutating viruses. @RecursionPharma / @valence_ai – AI drug discovery platform combining chemistry, biology and machine learning. @CausalyAI – AI that reads scientific papers and maps biomedical knowledge graphs. Fractile – building new AI chip architecture. fractile.ai @mindgard – "red teaming" AI models to find vulnerabilities. @Limbic_ai – mental health triage AI used in NHS services. @LindusHealth – clinical trials run with AI infrastructure. Anima Health – AI triage platform for healthcare providers. animahealth.com AUAR – robots that build buildings. AI for automated construction. auar.io @testudoglobal – insurance for the AI economy. Isomorphic Labs – DeepMind spinout doing AI drug discovery. isomorphiclabs.com BenevolentAI – one of the earliest AI drug discovery companies. benevolent.com London AI ecosystem is weirder than people think. Still mapping more. Comment any i've missed so far. londonmaxxxing.com

@burkov I suspect most people currently use it for their own personal use - not really for releasing. In my case I’ve set it up to sync personal emails, run topic models on them and manage them more easily. But the simplicity and flexibility for your own stuff seems to be the benefit.

@burkov While you’re right that all the components of OpenClaw were there already, what it did uniquely was have a clean architectural design of the gateway, channels etc that allowed people and agents to quickly build lots of bespoke stuff.

I didn't want to comment on OpenClaw. Usually, when there's so much noise in the media, it's some ordinary stuff just hyped well. So I took time to learn how it works thanks to open source. I was right. OpenClaw is 2% of ordinary stuff and 98% of hype. To put it very shortly, in case you were wondering, there are two things in it: 1. You can chat with an LLM via a text messenger. Not anything new. 2. The LLM can use tools that run on your computer. Not anything new either. Most of the "magic" mentioned in the media is about its ability to use the browser. But it's not *its* ability. It's Playwright's ability. Playwright is a library made by Microsoft which allows you to programmatically run a browser. It uses a built-in vision model made by Microsoft that converts the browser's screen into a textual description for LLMs. Again, Microsoft has built Playwright exactly for what OpenClaw is using it. So, OpenClaw's typical workflow: 1. The user types in a text messenger "Buy me a flashlight on Amazon." 2. OpenClaw blindly dispatches this message to an LLM which has access to some tools, including Playwright. 3. The LLM, trained not by OpenClaw folks, decides that Playwright is the right tool (of course it is) and Amazon is the URL to navigate to. 4. Playwright, built not by OpenClaw folks, runs the browser, which navigates to Amazon, and returns the textual description of what Amazon's home page looks like. 5. OpenClaw blindly returns to the LLM this textual description. 6. The LLM (again without any help from OpenClaw) decides that one should type "flashlight" into the search field and press Search, so it calls the Playwright tool with the search parameters. 7. OpenClaw calls Playwright because the LLM told it to and types "flashlight" and then presses Search (it's all part of what Playwright does out of the box). ... In the end of this LLM-controlled scenario, the order is submitted. OpenClaw just listened to what the LLM told it to do via tool calls. I tried hard, and I haven't found anything else worth mentioning in the source code. There's also a part that keeps "memories" about past conversations, but it's all basic stuff. These memories are stored in text files and grep (controlled by LLMs trained to use grep, and trained not by OpenClaw folks) is used to search in them. It's a nice hobby project, just like Cursor or Perplexity are nice hobby projects, but there's nothing there to look for, except for the hype and 2% of unoriginal plumbing code.

Regardless of whether a wealth tax is a good idea, the process seems broken.

Not really democratic any more - the AG can single handedly force people to move if there’s any risk the population might vote something through. Critically - they have to make the decision before they know whether there is a democratic mandate.

The retroactive part of California’s Billionaire Wealth tax is highlighting a major flaw in the initiative process for adding laws to the ballot: the Attorney General plus one person can basically force massive life choices without a democratic mandate Here’s how it works…

@agupta A lot of the issues with automation are down to the signaling systems deciding if you can go on a section of track. Apparently there are parts of the Uk network that still require Victorian semaphore systems.

It’s absurd that we’re going to have universal self driving cars but somehow trains that have a single degree of freedom require not one but two operators. Train agencies regularly run fewer trains than they have because of staffing requirements like this. Unsurprising bc this is a job no one should want. To everyone who complains about the low frequency on the T — this is why. We have the trains we don’t have the drivers. Automate the trains, run em at 1 min frequency, build more dense transit connected housing, and everyone will benefit, including workers from the abundance of new opportunities that will emerge. Some of the most labor-friendly places in the world already do this eg Paris. this is not rocket science.

We’re excited to announce SIMA — a smarter, faster way to log surgical cases and track training progress. We’re starting with ophthalmology and going live mid-July. Check it out! simasurgery.com

@FaresAntaki Nice idea! Hope it goes well!!

I started my first company. SIMA is a platform to make surgical case logging and progress tracking easier for trainees and programs.

@bumbadum14 Interesting to see the research in this! Seems this has changed over the last few decades. Used to be more women but increasingly more men.

My 9AM catholic church has exactly 0 obviously single women in it. The young adult group had 0 single women in it. When I was attending a protestant church before my conversion had exactly 0 women in their young adult ministry. You are not living in reality.

Fun talking today about data sovereignty within clinical trials in LatAm with @Farma_con. Seems like LatAm has a lot of interest in expanding what they do in clinical trials and does seem AI could play a big part!

The main benefits of the pure python approach came down to data augmentation and being able to change the learning rate for different layers. Those could obviously be added to the no code interface, but would mean a more complex experience. 3/4

Come visit us during the poster sessions if you want to hear more (or later in the week at the Bitfount booth). 4/4

Lots of fun presenting our experiments with clinicians doing no code fine tuning using Bitfount, at ARVO with Tahm Spitznagel, Gabor Somfai and team 🧵1/4

@jefielding Don’t know if acqui-hire has to be a win, but the American culture of being positive about the attempt even when a company fails is a big part of why America has had more successful startups than Europe imo

If your startup fails, all investor money is lost but a soft landing / acqui-hire is secured by the team, should you celebrate it all over social media as if it’s a win? I find this practice in our industry so weird, so common and everyone just goes along with it 🤷🏼‍♀️

@HawleyMO Property rights are too weak for personal data imo. Need rights which force the ability of individuals to take data relating to them, even if they “sold” it

How about we delete IP law for big corporations and protect property rights in personal data for everyday citizens