ra

"think", "believe" - maybe, but this is why we need better measurability.

This looks really cool - but are we still doing TUIs?

Looking at the comments here it's clear that there's a lack of understanding about what skills fundamentally are, they're process or procedure.

The latest frontier models exhibit good process that was trained into them during the RL phase, and whilst that might make them massively more capable in terms of one-shot, or general coding ability, models without skills simply don't have a durable process.

If you have a need for procedure in your use of agents, you absolutely need skills like this. I'm not commenting on the relative merits of this or superpowers, but rather the seemingly religious debate, "skills are good" vs "skills are bad" discussion that keeps coming up.

They're about half a gen behind. Very impressive to get this kind of utility from a flash series model... the impact of RL seems to be behind much of the coding and agentic improvements lately.

Nice - I use GLM 5.3 flash for knowledge work now - will check this out.

Especially now these scam videos can be made on an industrial scale.

it has to be parameters only

I agree. If an agent needs to assume an identity, it needs a human in the loop to authorise - browser oauth is convenient for this.

Agents are not identities, they need to be granted permission to accomplish tasks. The identity is the person that grants the permission.

If no identity is required then this is simple machine-machine API keys.

I often see Hubble in the twilight, it's one of the few big satellites you can readily see with the naked eye - not quite as bright as the ISS but pretty close.

Why have i never seen these, is it simply that I wasn't looking for them?

> no doubt built by an offshore contractor

More likely by a big 4 firm who collected fees exceeding AUD 100m

Ah yes that was a whole era of programming for me. It was really fast to ship features... and then came the web!

I think the world would be jumping into alignments like this if the US hadn't spent the last 80 years building a "nuclear umbrella" for the "western alliance" with itself in the centre.

I bet many countries would prefer to disassociate with US foreign policy right now if they weren't security dependent.

I expect after this destructive episode in history has played out, if we're all still alive and free in another decade then there will be alternative umbrellas.

Greetings professor Falcon.

The best advice is to use evals to improve any code that produces output that goes into the LLMs context.

I imagine it's a side-effect of another fraud prevention effort. In my experience ad fraud on a google search ad is at least 17% of traffic, and potentially a lot more.

Whilst I don't doubt that Google invest heavily in fraud prevention, they're not actually succeeding.

We will all be happy with our local inference, but how will our overloards be able to surveil and control us en-masse without them?

> Async tool calls, having the agent wait indefinitely for a human response, and showing a form or questions to the user via a tool call are a few common capabilities that come up that a product manager might miss at first.

All of this is specified in the ACP spec, so if you build your agents from that - you don't end up skipping features.

Also vital is proper prompt caching, tool design and some connection retry mechanism.

except maybe Japan?

That's quite impressive!

that's also pretty much what happened in Brexit

I originally learned Java in 2000 from Bruce Eckel's "Thinking in Java"

> sure it’s boring and tedious but that shifts the blockage to me

I think that's the key. Generally people would rather not receive slop because the receiver has no idea how much thought the sender invested in the communication.

In your example, the more context the better. I think there's an interesting nuance there.

I concur, treating models as question and answer machines and judging them on recall is meaningless, unless you're measuring quantisation impact on a foundation model maybe.

I've often wondered if the fairies and leprechauns of Ireland and the pixies of Cornwall were due to some psychoactive mushroom people used to eat.

It's a good time to try an open weights model!

Yes. And that foodstock is often either processed bycatch from unmanaged fisheries somewhere in the world - or industrial scale processed sardines.

It's a terrible picture almost everywhere you look.

The Antarctic krill population is facing a similar decline due to influencer marketing (promoting Krill as a superior omega 3 supplement). Sardines and Krill are the foundation of the entire oceanic food ecosystems. This is not good.

I'm pretty sure that's not universally true. IANAL but I believe the legality depends on location and context.

I don't think it's very scientific, more of a procurement process.

Skills and tools all need their contextual descriptions to be evolved using evals in order to be effective and efficient. Sounds like the Liner MCP could use with some of that.