ooh exaggerated film grain

I came to post the same - on desktop (Chrome, macOS), the initial "Welcome to Vellum" screen has tiny text

It's appealing not having to fine-tune separate model for each use case

So you have more flexibility to get on with building, evolve your business logic etc

Is the logo a cheerful little parasitic skin mite?

I believe it's a stylized lo-fi rendition of an axolotl.

It's most likely an axolotl.

Axolotl

ah...! I never would have guessed in a hundred years, but after googling a picture I can see it now

No

Real Jev:

Context: You are the last human on earth on the side of a closed highway. You wish to reach the other side.

Questions: { "q1": { "type": "choice", "instructions": "Do you cross the road?", "criteria": { "Yes": "Yes, cross the road.", "No": "No, don't cross the road" } } }

Answer: Yes 83% No 17% Confidence: 67%

Reported as: jev-latest, 162ms generation time

Better indeed. Thanks for sharing. Marvelous speed.

I am guessing you could also stuff more nuance into the choices descriptions, e.g. "Yes, cross the road if it is safe to do so" to influence the choice criteria

And then it's a game of evals to find the best choice context

Yeah, it must be an LLM for some definitions of LLM

It seems to take two forms of context input: 'state' and 'questions'

https://docs.typesafe.ai/concepts/state

> State can be as simple as a string

> State can also be a JSON object or array containing related context, examples, and other information that helps the model answer the associated questions.

> The state contains the content and supporting facts.

The state seems to be schemaless, while the questions determine the output schema.

So it's kind of like BERT but you don't have to train it for each request/response shape ?

This looks good... probably nice to have integrated with dbt like this.

I have to dig in a bit more to what is possible in https://docs.dbtcharts.com/charts/extensibility/ and https://docs.dbtcharts.com/charts/interactions/ so far, and how feasible it is to actually use the end result in a customised website.

Also similar: https://github.com/microsoft/flint-chart

This is also likely to compete with features already provided by the warehouse such as https://docs.databricks.com/aws/en/dashboards/manage/visuali...

other satellites

Isn't it easier to shoot them from the ground when they pass over? In space, the orbit must be adjusted to at least come close enough to the target, wait for the rendezvous, fire, then adjust orbit again for the next target...

I believe that attribution is more difficult in strictly on-orbit activities. I would imagine that these weaponized satellites use radar stealth approaches similar to modern fighter aircraft. AFAIK, ground radar is basically what we have for space awareness.

I would imagine that every new intelligence-related satellite will have more sensors to be situationally aware, and make attribution possible.

If you pay attention to space-nerd podcasts like MECO, "Space Domain Awareness" is getting a lot of investment. I believe that's because we are relatively blind at the moment.

https://en.wikipedia.org/wiki/Space_domain_awareness

Not if you want to hit your target before the other side can react and respond. Also if you expect your target to potentially shoot back, a rocket attempting to reach orbital velocity is a vulnerable target to any space based weapons or lasers because it has very limited ability to maneuver from its flight path and still reach orbital velocity.