ciyer

software engineer, seduced by data

[ciyer @ciyer]

I just published Web 3i https://t.co/9umTAvJu7U

[Prachee Avasthi @PracheeAC]

Love all the AI-native initiatives for research sharing I’m starting to see pop up across the ecosystem. If you’re building in this space, I’d love to hear from you as there may be some synergies or opportunities across our open science efforts at @AsteraInstitute

[Prachee Avasthi @PracheeAC]

We’re still looking for a lot of folks for Open Science at Radial (within the @AsteraInstitute ecosystem) including a data steward, operations associate, as well as an engineer and product manager for the Stacks (https://t.co/5DNdQoe4DA) to work on our publishing platform. If you think any of these roles are a good fit for you, apply now!

https://t.co/ylu3NBsMGs

[ciyer @ciyer]

@TheStalwart Also came away from the NYT article unconvinced. IMHO, this remains the most convincing dive into Satoshi real identity: https://t.co/XQJffKbOuO

[Becky Pferdehirt @beckypferdehirt]

@AsteraInstitute is launching an essay competition to identify systemic bottlenecks to today's science. We're looking to hear from currently practicing scientists first-hand: what is broken?
Deadline is May 1st. More below 👇

[ciyer @ciyer]

@TheStalwart That's wild! I love that network that includes "She Blinded Me with Science," "PacMan Fever," along with Vaggione and Mort Subotnik as nodes.

[ciyer @ciyer]

@TheStalwart How did you come across this piece?

[ciyer @ciyer]

@TheStalwart Vaggione is so awesome!

[ciyer @ciyer]

@adjajadikerta If I had the resources, I would love to run the experiment of training a model on all science up to 1904 and seeing if the model could be prompted to produce special relativity.

[ciyer @ciyer]

@adjajadikerta Great, thought-provoking article!

In Kuhn’s view, one paradigm is chosen over another because it better reflects the considerations a community of scientists has come to view as important. It might be necessary to model and optimize for social dynamics in a community.

[Astera Institute @AsteraInstitute]

We're expanding our life sciences work through a new division Radial led by @beckypferdehirt as CEO. Becky be work closely w @PracheeAC @seemaychou to help lead experiments with infrastructure and approaches to funding, enabling, and publishing science 1/

[ciyer @ciyer]

https://t.co/VmtqXxch9y

[ciyer @ciyer]

@srikosuri @supply_side_acc Been a while, but IIRC, Kuhn was more of a relativist, arguing that paradigms are incommensurable because they place value on different things, and notions of truth are paradigm dependent.

[ciyer @ciyer]

@typesfast Using the freezing and boiling points of water, a common and important compound, is a good idea, but as the Sumerians and Babylonians knew, 12 and 60 are much better bases for a numeral system.

[ciyer @ciyer]

@arpitrage Compute it analytically or simulate draws from the distribution… either way, you’ll see that though you will have more M overstock in absolute numbers, the overstock rate is much higher for XL.

[ciyer @ciyer]

@arpitrage You produce the E for each size: 100 M, 10 XL. You get a truncated binomial of sales (you cannot sell more than you made), and the expected overstock is number_produced - E[min(number_produced, sales)].

[ciyer @ciyer]

RIP Martin Parr https://t.co/xJhg0vpfcC

[ciyer @ciyer]

@trashpandaemoji You have convinced me. Yeah, REST with some description of the endpoints like Swagger, should be enough for remote resources, and --help should be enough for LLMs to interact with local endpoints. Not sure where MCP is needed anymore.

[ciyer @ciyer]

@trashpandaemoji To cover what MCP does, that would be enough. E.g., for a weather service, an API key + location gets you a weather report.
I was thinking further, though: what if I also had a cool visualization to embed? That would need to be formatted for the display in an LLM chat.

[ciyer @ciyer]

@trashpandaemoji For CLIs, I think you are right. For UIs with an inherent graphical component... not sure. But I could see just using HTTP/HTML(CSS/JS), looking at the user agent and embedding context using ARIA to optimize for LLM integration.