Leynos

7 ms·

61.
by Leynos 5mo ago
That is not auto-complete, that's API Usage Examples. See: https://marketplace.visualstudio.com/items?itemName=VisualSt...
62.
by Leynos 5mo ago
Yes, thanks. Too late to edit now, sadly.
63.
by Leynos 5mo ago
Is that taking cache hits into account?
64.
by Leynos 5mo ago
Not that I give much credence to anything Zitron says, but the amount of inference you can get on a £200 a month OpenAI or Anthropic subscription is easily an order of magnitude more than what you'd get paying the same amount at subscr
65.
by Leynos 6mo ago
Kimi 2.5 was like using Sonnet 4 on a flaky ADSL line. I haven't tried K2.6 yet, but the physical unreliability of the connection was too off-putting.
66.
by Leynos 6mo ago
Isn't that essentially GPT Pro Extended Thinking?
67.
by Leynos 6mo ago
Their top tier plan got a 3x limit boost. This has been the first week ever where I haven't run out of tokens.
68.
by Leynos 6mo ago
Combine that with emmet / zen coding: https://en.wikipedia.org/wiki/Emmet_%28software%29?wprov=sfl...
69.
by Leynos 6mo ago
OP should link to https://www.writetothem.com/
70.
by Leynos 6mo ago
How does this work? Someone names a coin after his project, pays him commission for every transaction so he will be incentivised to give it positive coverage on his blog, then sells their stake once the value inflates leaving bagholders wit
71.
by Leynos 6mo ago
That's pretty typical of 404 Media's reporting style.
72.
by Leynos 6mo ago
From the article, here is the recording of the meeting: https://www.youtube.com/live/xLPF3rTT0mY The event in question can be seen around 1:56:00 through 2:00:00 (approx)
73.
by Leynos 6mo ago
It amazed me that no one picked up on Codex last summer when it was effectively unlimited. I must have burnt through £10k worth of inference whilst still paying £20 a month
74.
by Leynos 6mo ago
I just used it because I preferred the UX.
75.
by Leynos 6mo ago
They got rid of "nerdy" last month too. Shame
76.
by Leynos 6mo ago
How well do LoRAs work for this using something like Thinking Machine's Tinker?
77.
by Leynos 6mo ago
Also, fuzz your benchmarks
78.
by Leynos 6mo ago
Opus 4.6 currently leads the remote labor index at 4.17. GPT-5.4 isn't measured on that one though: https://www.remotelabor.ai/ GPT 5.4 Pro leads Frontier Maths Tier 4 at 35%: https://epoch.ai/benchmark
79.
by Leynos 6mo ago
It's not difficult at all to burn through your weekly limit just writing code.
80.
by Leynos 6mo ago
Wondered if you were able to suggest what might be stopping large scale build out of sodium ion and redox flow batteries?
81.
by Leynos 6mo ago
You just said it. If consistency is that important, keep consistent versions of model, harness, prompts, skills, etc., and regression test changes. That way lies madness :)
82.
by Leynos 7mo ago
It feels to me that "walk down the design tree" has a specific meaning with respect to treating the design as a hierarchy (although whether that means BFS or DFS is still ambiguous). "Be critical" lacks that specificity.
83.
by Leynos 7mo ago
Use evals Coming soon, unit, behavioural and regression tests for your prompts and skills :P
84.
by Leynos 7mo ago
Kimi is surprisingly good at Rust.
85.
by Leynos 7mo ago
Again, here's what works for me. When I get an idea for something I want to build, I will usually spend time talking to ChatGPT about it. I'll request deep research on existing implementations, relevant technologies and algorithms
86.
by Leynos 7mo ago
And do you have statistical evidence to back up your claim of increased enforcement, or are you just reading about it in the Daily Mail?
87.
by Leynos 7mo ago
Yup
88.
by Leynos 7mo ago
Here's what I suggest: Serious planning. The plans should include constraints, scope, escalation criteria, completion criteria, test and documentation plan. Enforce single responsibility, cqrs, domain segregation, etc. Make the code as
89.
by Leynos 7mo ago
Okay, here goes. You can tell when someone is acting in bad faith when they talk about a law that has been in force and enforced since the 1960s is something new. Of course, "touch grass" works just as well.
90.
by Leynos 7mo ago
It's fun. That's the utility. Although you could also probably convey information in a visual shorthand in a way that is quicker to parse than a table. I want one that shows my agents working in a hipster coffee shop with chrono t

More ›