DenisM

5 ms·

1.
by DenisM 7d ago
Or better context curation - less lossy compression saving back to context. Maybe even jettisoning part context into an external semantic store instead of conpression. Or placing less data into context to start with. Or a combination of all
2.
by DenisM 7d ago
Great story! Verifying sources is a recursive problem - where do you stop? Humans have intuitive feel for it, but agents don’t or at least not yet (I wonder if intuition is just a secondary neural net which is currently being added to the a
3.
by DenisM 7d ago
I observed the same, and generalized it as inability to recognize salience and more broadly apply discretion. In turn it makes me wonder how do humans do those things? Perhaps it is our human job to apply discretion going forward.
4.
by DenisM 7d ago
Probably a quote from 3-body problem.
5.
by DenisM 9d ago
The Trump and former president terms were likely firmly stuck together in the embedding space. The model doesn’t validate every single token it produces because validation itself requires tokens. A bloom filter of outdated embeddings will
6.
by DenisM 9d ago
How so?
7.
by DenisM 9d ago
It’s probably brittle though? Replication implementation has to change in some ways from one version to another.
8.
by DenisM 12d ago
I like to think the real world lessons in failures are valuable. If I have this idea one day I will search for it and then think “how am I different fro that which already failed?”.
9.
by DenisM 12d ago
You may be interested in TITANS: Test-Time Learning: The model updates its own memory weights while running an inference task.
10.
by DenisM 13d ago
Perhaps our own statefullness is a hack of nature. We have electrical signals in our brains, neurotransmitters, neuron growth. By any reasonable measure it’s a hack on top of a hack. But it works well enough for us to get buy. So it does f
11.
by DenisM 14d ago
For practical fixes, consider replacing the TV with a “commercial monitor” and connecting an Apple TV device instead (or an open source thingy). Disconnecting “smart tv” from internet to use Apple TV is possible and might work, but you will
12.
by DenisM 17d ago
The original selling point of mobile web (tiny screen with text-only data, before real mobile web) on mobile phones or even watches was checking stock prices and weather. It was really weird that checking stock prices was something you need
13.
by DenisM 19d ago
Most valuable thinking is done at the margins, where you don’t have much capacity to emphasize with a diverse and unknown audience. That said, abdicating to an LLM is the worst of all worlds - you’re not thinking and the product is not tail
14.
by DenisM 20d ago
In very broad strokes, it seems that Europeans on the average have much more of these things than Americans. Are they noticeable happier? Would be nice to hear from someone who lived both sides of the pond for a long while.
15.
by DenisM 21d ago
Does it work? Journaling is a major commitment, I hope it pays for itself.
16.
by DenisM 24d ago
Do you find that given a formal spec an agent can write complete implementation you don’t have to even read? I keep thinking about various ways of “pushing back” on an agent, shortening feedback loop and extending what we can grantee about
17.
by DenisM 24d ago
Is the source material good, in your opinion?
18.
by DenisM 26d ago
Can it debug the apps? That would the app singularity - user speaking at their phone until phone complies and produces desired app for the current moment.
19.
by DenisM 27d ago
Memory-hard hash functions maybe? Like, you must dedicate 4gb of ram to compute the function. Not a problem for a one-off, but is a problem when reading lots of pages at once. Or… the site will serve a random seed and the device must comput
20.
by DenisM 29d ago
Or calling into a full-knowledge model “I’m facing problem x, how do I ask myself the right questions?” I should do that myself, come think of it.
21.
by DenisM 29d ago
I think the labs will always get the first stab at breaking the vm, and fixing it, while developing newer models. Whatever people can do later is what labs already did.
22.
by DenisM 1mo ago
Do you explicitly ask for metaphors? I wonder if we need a list of things to tell AI to stay sharp, like this one. Sometimes I tell the model existing design is stupid and then it explains reasoning to me.
23.
by DenisM 1mo ago
I’m guessing the new world will be a small set of VM tech that’s consistently hardened by all labs every day with each new model before model release. This won’t make the tech secure, but it will nullify models ability to breakout by making
24.
by DenisM 1mo ago
> They're fast and deterministic and I run them in git pre-commit. Isnt that too late? I would want the agent to stumble into this as early as possible in the agentic loop, eg at the same time as compiler.
25.
by DenisM 1mo ago
Once upon a time my grand-grand-boss explained that he had no problem hiring top-notch engineers - pick up the phone, call recruiters for new candidates, connect the incoming stream to the interview pipeline, two months later you get the re
26.
by DenisM 1mo ago
Does CTE not allow composability?
27.
by DenisM 1mo ago
Snowflake has STREAM which can be crated on a view. It also has Dynamic Tables, from which you read a delta using STREAM or using row timestamp. SQL server has Query Notification. You can also read from debezium or other cdc, but thats more
28.
by DenisM 1mo ago
Fair point. I’m in a position of “creating horizontal influence” and I’m lucky enough to be able to focus on less than a dozen outcomes per quarter, so that colors my perception.
29.
by DenisM 1mo ago
You missed my point. As a possible reader of your email I’m facing dozens of emails every day, my default reaction is to skip past and “read it later”. Maybe you did put a lot thought into it, how would I ever know? If you show up in person
30.
by DenisM 1mo ago
> what have you found so far? Love it, thank you!

More ›