DenisM

5 ms·

31.
by DenisM 1mo ago
> talk, don’t write” people is based on their inability to logically think through, structure, and communicate their ideas and, instead, would rather connect directly and hope the solution emerges from a conversation. That is a remarkabl
32.
by DenisM 1mo ago
Whats the worst that could happen if you read GP’s words in a positive bias? Best case you learn something, made a friend, amd trained yourself to look for useful feedback in the future. Worst case you improved your mood and moved on. You’r
33.
by DenisM 1mo ago
I actually got repulsed by it, and now I’m avoiding ChatGPT. I can’t easily explain why.
34.
by DenisM 1mo ago
> as the LLM progressed through my project, learning more and more about it… This means that it needs to start over, learning the project. I am in the process of exporting from the previous account, but I’m not sure how well this will tr
35.
by DenisM 1mo ago
When you get too nostalgic just think back to XSD type system in Xquery. That should cure the nostalgia real fast.
36.
by DenisM 1mo ago
The article really just glosses over this. Just because a model doesn’t kno a fact does not mean it will go and fetch it, the model just as well can invent the fact. For that to not happen the model needs to know that it’s missing the infor
37.
by DenisM 1mo ago
Shuttling? Once you remove the motor and put it into a backpack downhill riding should be quite decent. Decent enough I think to make it worth it for the space and cash saved.
38.
by DenisM 1mo ago
Mountain bikes one level down from professional tier start at 6-8k. If I could pay 2k and get a nice electric bike from my nice muscle bike it would be a bargain. However bimotal is not that nice - adding a motor to rear triangle alters wei
39.
by DenisM 1mo ago
Sometimes it’s just an awkward turn of phrase on my part that creates a wrinkle in the conversation. Sometimes agents identify that, and we can work together and direct that, but correction itself eventually loses competition to the origina
40.
by DenisM 1mo ago
Sometimes I ask the agent why it gave a certain answer, when I feel it overly fixated on something. It would be cool if the ui hilighted the poisonous part of the conversation somehow.
41.
by DenisM 1mo ago
eventually we will get to “just use postrges” stage
42.
by DenisM 2mo ago
I wish they don’t stop. Make it easier to skip past.
43.
by DenisM 2mo ago
Double-harnessing distortions seems like a weak point. I wonder if it’s just a temporary measure and long term it’s about writing custom harness going straight to inference APIs across all models.
44.
by DenisM 2mo ago
+1 well written, well paced article. Pleasure to read. Have you tried measuring Gemini? now that you have the router it should be a simple task. Thanks!
45.
by DenisM 2mo ago
How can Smart Router achieve higher task completion rate compared to any of the base models if all it does is dynamically switch base models based on cost??
46.
by DenisM 2mo ago
Years ago I learned that GM’s financial arm made more money than the car company. The other day I also read that airlines make more money from selling points than from flying people. Crazy world.
47.
by DenisM 2mo ago
This might grow bigger than cobol. There will be a large number of companies that will be staffed entirely by vibe coders and when an issue crops up that cannot be debugged by them a consultancy will be brought it. The consultants will use
48.
by DenisM 2mo ago
I kinda dislike where your comment leads, but I can’t argue with the logic either. Time for self-reflection I suppose. Maybe a better way of doing this is to ask the experts to do the work? But then again, someone who does not understand th
49.
by DenisM 2mo ago
Diabolical. I’m not sure if I’m frightened or fascinated… , I’m thinking of where can this lead if applied iteratively. Like, slop supernova followed by by a slop brown dwarf?
50.
by DenisM 2mo ago
Suggestions? My attempts to get the agent to summarize the discussion largely failed. Other options I’m considering is capturing design, implementation plans, and maybe sub-plans. I’m Other words precursor to actions. Another idea is to get
51.
by DenisM 2mo ago
So what is tre answer then? I feel like “discretion” parties missing. Do we need some kind of special discretion model?
52.
by DenisM 2mo ago
Not quite a Bear Sterns scale event.
53.
by DenisM 2mo ago
How do you know what you built is good? That was the pivot for me - I defined it for myself as it’s something that makes a positive contribution to my users.
54.
by DenisM 2mo ago
For clarity: Indian government.
55.
by DenisM 2mo ago
Remarkably, it took non-AI workloads 25 years to grow from zero to 3%, and it took AI only ~ 2.5 years to get to 2%. This puts a number to my intuitive feeling that things move an order of magnitude faster this time.
56.
by DenisM 2mo ago
Where did the tokens go?
57.
by DenisM 2mo ago
> So what about the other 3%? That's not "massive"? Or it gets a pass because it's not AI? How did you read that in my comment?
58.
by DenisM 2mo ago
Select quote: * 5% of electricity in the United States is used to power data centers. For AI-focused ones specifically, it’s probably around 2%.* I think it’s massive.
59.
by DenisM 2mo ago
This media frenzy can go on forever tho. A Bear Sterns moment would be more solid. Oracle might ge the first to collapse if things go south, so we might be fine until then(?).
60.
by DenisM 2mo ago
In my little ai bubble the original blog post was very influential. It does sound immodest, but could also be true.

More ›