tudorizer

Why the downvotes?

Has anyone pointed CC at Pharo? Or is the UI and IDE a limit in this case?

deep research

This must be one of the best and most widely impactful programs for kids I've heard about recently. Such a different tone than all the other nonsense happening around.

How are above-average tech people still working for and listening to this bloke? What am I missing?

I'm curious about the "The irony is not lost on me that when I presented Claude with the findings of our week's work, it immediately flagged our account with Anthropic." part. Why would they flag your tests?

That's to be expected from specs alone, no? The memory bandwidth is the culprit. The comparison is almost unfair for most inference, especially on a dense model.

I wrote more details about what's appropriate for a DGX Spark here: https://spark.enverge.ai/blog/dgx-spark-prefill-vs-decode

[disclosure: co-founder of the company linked]

https://spark.enverge.ai/ a growing platform of nano data centres, focused on "the rest of us". Very light-weight virtualisation layer, great for Grace-Blackwell experiments and general benchmarking.

huh!? I entered this text in the submission form, not as a reply.

Spark Studio: The Dashboard That Fixes Its Own Bugs

They've gone up in price a bit now, just like most of the market.

Still a fair option for renting a DGX Spark.

https://lab.enverge.ai/b200-challenge/ - making high-end compute available to a few researchers, who are conscious about the impact of data centers on climate and local communities.

I'm working on something that proves to be more ambitious than initially thought: pure Python notebooks + zero-emission GPUs + git = https://lab.enverge.ai

Currently struggling with an experiment where DeepSeek-R1 is being overly verbose.

There's a smell of pseudo-science here. That weird blend of interesting + plausible with sprinkles of heavy-handed parallels.

After "Isaac Newton, who may have been the smartest person who ever lived" the level of trust fell drastically.

Sure, the periodic table was extremely useful and we were using electricity before we understood it, but we understand LLMs far better, mostly because they are our own creation.

Maybe the lines between exploration, creation and discovery are fuzzy sometimes, but this article tips over into AI propaganda.

Well ... what is independent analysis? The author is not a reporter, but someone who understands business principles.

I think you got the causality the other way around here.

Right, but critiquing with the right perspective is important. Statements about making a loss must contain the entire economic picture, otherwise they simply aren't true at some point.

A business can't be scrutinized unless the units of economics are understood.

Most likely tickling an own itch. Also to validate/invalidate if sanity wasn't lost.

Plus, Ed's articles have been circulated in some investment groups and nobody expressed a clear counter-point.

PS. I wasn't familiar with that saying.

> because the average person is bad at explaining what they want an LLM to do

Agreed. It's the saving grace for most platform which integrate LLMs even right now. Eg. v0 narrows the scope of general purpose LLMs and offers educated guides.

This video nerd-sniped me so hard. All these mechanics put a smile on my face.

> Boundaries and purpose

This can't be emphasized enough!

As long as you're not going too far down the "the hammer will start using itself" path.

Recursive feedback loops and fast pace of improvements are priced in.

This hammers starts sounding like Excalibur given all the mythos around it. :)

Agreed on the minimal-chargeable-bar point. The beauty arises when juniors grow.

I have tried quite a few and would defo agree with your conclusion. Also not fully buying the "you're holding it wrong" answers.

I point at long such lists of tools only to indicate a certain level of complexity, which will most likely fall in the realm of "oh, this is too technical for me. I should delegate this to Alice, because she good with tech". This is only shifting the problems and problems mean opportunity.

Absolutely.

When complexity grows and lines between boundries of what's what blur, opportunity for misunderstanding sneaks in.

We should welcome scrutiny, though.