wolttam

10 ms·

151.
by wolttam 4mo ago
95/4/1 holds here too
152.
by wolttam 4mo ago
Depending on who you are and how you use these models, we're already at this point
153.
by wolttam 4mo ago
I'd agree that it has gotten better over time. I'm glad we have Nextcloud around.
154.
by wolttam 4mo ago
I'd love to know of such an example where a U.S. LLM blatantly denies something factual. Maybe I'm living under a rock but I can't think of one
155.
by wolttam 4mo ago
I rarely work on anything that demands better than DSv4 Flash, let alone pro. If I can describe the problem and its solution well enough, Flash just does it. If I can’t (or am feeling too lazy to) describe the problem well enough, and can
156.
by wolttam 4mo ago
I’d be willing to bet that most of us can see both the positives and the negatives of AI just fine. It’s about the greyest subject I can think of.
157.
by wolttam 4mo ago
That's... $12.50/hr per GPU. WTF?
158.
by wolttam 4mo ago
As a human, I'd rather just explore the code (which I tend to trust more than anything written beside it) myself. The bot though- I don't want it to waste a bunch of tokens exploring nonsense. If I can write a few lines of text th
159.
by wolttam 5mo ago
I think the idea is that the model is seeing embeddings that map directly to underlying pixel data, rather than being fed semantically rich embeddings from an encoder model which itself had seen the raw pixel data.
160.
by wolttam 5mo ago
The low contrast of this website hurts my eyeball
161.
by wolttam 5mo ago
Because they like and are excited about Zig.
162.
by wolttam 5mo ago
MCP and shell/bash tool calling serve totally different use-cases, this discussion is... odd.
163.
by wolttam 5mo ago
At this point in time I doubt that AI is ever going to escape the control of the humans with capital, and those humans will just use it as a tool in order to amass ever more capital. The hopeful outlook is that AI is also a lever for the wo
164.
by wolttam 5mo ago
Businesses will still be owned or at heavily influenced by individuals and it is those individuals that will be able to steer how those business’ capital is allocated (including towards themselves).
165.
by wolttam 5mo ago
I think machines like the DGX Spark are about to become a lot more common/popular. It’s big enough to run sparse 150-250B MoEs with enough throughout for a single user. Deepseek v4 Flash is #1 (in terms of usage) on OpenRouter because
166.
by wolttam 5mo ago
Pretty much every durable system has an intent log of some sort. The log provides durability, the database system just integrates that log into a more queryable format.
167.
by wolttam 5mo ago
“Using the agent after trying for 20 minutes” This made me chuckle a bit. There’s no point in fooling ourselves about our own skill retention if this is the case.
168.
by wolttam 5mo ago
A big point of DeepSeek V4 is the significantly reduced KV cache size.
169.
by wolttam 5mo ago
You can check my profile for which one I like most :) I do think there have been efforts to benchmark different harnesses. Personally I'm not going to choose one harness or another based on +/- a few percentage points in a benchma
170.
by wolttam 5mo ago
Probably a matter of taste. I prefer the harness I wrote, I don't want to go near Anthropic's bloated mess of a harness with a 10-meter pole.
171.
by wolttam 5mo ago
I've noticed that models have gotten less finicky with this over time. Harnesses don't need to be complex to get good coding performance from models, they just need to implement some sane primitives for code exploration and editin
172.
by wolttam 5mo ago
The value of Claude Code the harness isn't that great. There's a lot of other good harnesses out there.
173.
by wolttam 5mo ago
I was hoping they were going to do this. I'll keep running Flash locally for the stuff I care about data privacy, but the value of Pro through their API is unreal for anything else (and I want to give them my training data as long as t
174.
by wolttam 5mo ago
Yes, but that tends to limit the reach of your content. Hence why a lot of people reach for ads. Between seeing ads and doing a little bit of proof-of-work for the author, I'd choose the latter.
175.
by wolttam 5mo ago
The idea would be that AI companies would still be forced to do this proof of work. Anubis proved the idea
176.
by wolttam 5mo ago
Yes, but: > Although Anubis could be altered to mine cryptocurrency to serve as proof of work, Iaso has rejected this idea: "I don't want to touch cryptocurrency with a 20 foot pole." Which in my mind is a shame. Crypto is
177.
by wolttam 5mo ago
I’ve been thinking of a proof-of-work scheme for accessing content where you effectively need to mine some crypto for the author, but, this idea might not fly today
178.
by wolttam 5mo ago
Frontier: Need it done quick and I'm willing to pay. Open-weight: Good enough for the majority of tasks, and I'm willing to spend a bit more time and effort steering towards my desired result.
179.
by wolttam 5mo ago
It depends on the use-case. yes, 90% of cost is cache in agentic coding scenarios (actually 95% in my experience). But not when the model reasons for 200k+ tokens before answering a complex problem.
180.
by wolttam 5mo ago
I feel like my comment was either misconstrued or misunderstood and don’t have anything to respond after the inflammation at the end of your comment there. For the record, I don’t know anything about you or your kids.

More ›