Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
mwigdahl
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
5 ms
·
31.
▲
by
mwigdahl
3mo ago
The fully loaded cost of employees at that salary level is likely over $100K. You may still be right on the relative expenses, but you can't just look at the top-level salary line.
32.
▲
by
mwigdahl
3mo ago
Sure, but the original comment said "living out" those fantasies.
33.
▲
by
mwigdahl
3mo ago
I don't know -- it's not even showing Fable in /usage any more, in sessions that were using it just fine moments before. Seems more like a backend change to me.
34.
▲
by
mwigdahl
3mo ago
You missed a great opportunity to title this "The 10 most popular numbers in Hacker News titles"!
35.
▲
by
mwigdahl
3mo ago
For sure! It's possible there are relatively invariant properties that make one language or another "better" for LLM coding in general though (even if that's just the self-fulfilling prophecy of size of the training cor
36.
▲
Ask HN: Any objective research on which languages are best for AI agents?
5 points
by
mwigdahl
3mo ago
|
6 comments
37.
▲
by
mwigdahl
3mo ago
“An honest caveat…”
38.
▲
by
mwigdahl
3mo ago
There were a lot of European wars between the Napoleonic Wars and the First World War. I agree with your point overall but there was a lot of fighting each other during that timeframe.
39.
▲
by
mwigdahl
3mo ago
Yes. It is great for analysis, debugging, and understanding larger segments of code. I'll use it sparingly since it won't be on subscription, but there are cases where it will make sense to use it.
40.
▲
by
mwigdahl
3mo ago
I'm reading your story now -- it's great space adventure! Congratulations!
41.
▲
by
mwigdahl
3mo ago
Apparently Fable got all the power when writing this...
42.
▲
by
mwigdahl
3mo ago
> Fable returning wrong answers if it suspects the topic is sensative Citation, please? On the original release they downgraded models silently to Opus 4.8 when they suspected it was being used for LLM development, but they stopped doi
43.
▲
by
mwigdahl
3mo ago
Unfortunately as cost reduction trends to 100%, it comes along with an intrinsic high-pass sarcasm filter.
44.
▲
by
mwigdahl
3mo ago
Same (including the overthinking issue).
45.
▲
by
mwigdahl
3mo ago
Hence the fading away of the "nerd with pocket protector" stereotype...
46.
▲
by
mwigdahl
3mo ago
During the initial release, they indicated you'd be able to use Fable 5 as part of your subscription for a limited period of time, and then it would require usage credits. They also did say at that time that they hoped to make it a pa
47.
▲
by
mwigdahl
3mo ago
Sure, and if they hadn't trimmed logs someone would also be complaining that "stupid Anthropic lets logs fill your hard drive"! Oh wait, they just complained about Codex doing this a day or so ago.
48.
▲
by
mwigdahl
3mo ago
It's a good question, but for multiturn conversations even cached context adds up quickly. My experience has been that spawning off subagents for defined tasks in a large overall plan generally makes me come out ahead. I'm sure f
49.
▲
by
mwigdahl
3mo ago
I don't think all the negativity is from crazies, but big chunks of it are certainly motivated. I certainly left out numerous other categories.
50.
▲
by
mwigdahl
3mo ago
For agentic computer use Sonnet 5 low performs better than Sonnet 4.6 medium at just under half the cost, and better than Opus 4.8 low at 25% off. Their success rates are not that far off. Agentic search is a different story, but even ther
51.
▲
by
mwigdahl
3mo ago
Seems like a combination of multiple factors: "They took my shit away!" -- 3-day Fable 5 addicts (me) "How dare they tell Trump no?" -- US nationalist / "my country right or wrong" types "Great to see
52.
▲
by
mwigdahl
3mo ago
What? If you're comparing their models in the same size class, Sonnet 5 is Pareto-optimal over Sonnet 4.6.
53.
▲
by
mwigdahl
3mo ago
Not to sound like an LLM, but that seems exactly right to me. Use it as a cheaper, high-functioning task subagent and lower reasoning for a master Opus session. As long as not every portion of your task requires maximum intelligence, you
54.
▲
by
mwigdahl
3mo ago
Was that tech guy a crypto miner by chance?
55.
▲
by
mwigdahl
3mo ago
"Evil Anthropic implements rolling log files." Film at 11.
56.
▲
by
mwigdahl
3mo ago
Except that those customers can access the traces for 30 days, and freely copy them at any point during that period? It's a usability issue at worst, not some anti-customer conspiracy.
57.
▲
by
mwigdahl
3mo ago
Wow, that's fascinating! What else can you tell us about the boat? How big was it? What was the rigging? How long did they take on the circumnavigation? I've always dreamed of doing something like that.
58.
▲
by
mwigdahl
3mo ago
Jury time is paid time off at my US company. And at least at the US Federal level there's a daily stipend for sitting on one. Lower level courts may vary on that.
59.
▲
by
mwigdahl
4mo ago
I'll do you one better! Why is GLM 5.2?
60.
▲
by
mwigdahl
4mo ago
Article is clearly slop.
More ›