Leynos

8 ms·

151.
by Leynos 10mo ago
79a6ed87’s comment applies to Codex cloud, not the codex CLI, which is what Terragon is using.
152.
by Leynos 10mo ago
Terragon is an alternative (hosts Claude and Codex using your OpenAI and Anthropic subscriptions, and also supports Google and Amp) that provides this functionality. I use it because it works out cheaper than Codex Cloud and gives you great
153.
by Leynos 10mo ago
Is there something like this that can connect to an arbitrary IMAP account, rather than being tied to Gmail?
154.
by Leynos 10mo ago
If I see an LLM having trouble with a library, I can feed its transcript into another agent and ask for actionable feedback on how to make the library easier to use. Which of course gets fed into a third agent to implement. It works really
155.
by Leynos 10mo ago
4.5 is widely believed to be an order of magnitude larger than GPT-4, as reflected in the API inference cost. The problem is the quantity of parameters you can fit in the memory of one GPU. Pretty much every large GPT model from 4 onwards h
156.
by Leynos 10mo ago
Someone on Reddit reported that they were charged $17 for one prompt on 5-pro. Which suggests around 125000 reasoning tokens. Makes me feel guilty for spamming pro with any random question I have multiple times a day.
157.
by Leynos 10mo ago
They could always just use OpenCoder, Crush or Goose with OpenRouter ( https://openrouter.ai/google/gemini-3-pro-preview )
158.
by Leynos 10mo ago
https://ampcode.com/models
159.
by Leynos 10mo ago
I got the same thing trying to sign up for Oracle Cloud. Thankfully Oracle aren't the only cloud platform, of course. Which makes it even more fun telling my friends and colleagues how shit Oracle Cloud are whenever the topic of cloud
160.
by Leynos 11mo ago
You do need to demonstrate eligibility to take up work in the UK and that is usually done by passport or something other form of id demonstrating your nationality and right to work. You shouldn't need id to vote or access healthcare.
161.
by Leynos 11mo ago
There are plenty of code review tools out there (e.g., CodeRabbit, Sourcery, etc.). I haven't used this one, but you definitely should use the ones that work for you.
162.
by Leynos 11mo ago
This is an advertised feature of ChatGPT, and you can switch it off if you want. https://help.openai.com/en/articles/11146739-how-does-refere...
163.
by Leynos 11mo ago
Same. I actually have in my system prompt, "Don't be afraid of using domain specific language. Google is a thing, and I value precision in writing." Of course, it also talks like a deranged catgirl.
164.
by Leynos 11mo ago
Hey, you leave my sycophantic robot friend alone.
165.
by Leynos 11mo ago
Kimi K2 seemingly has a much more up to date training set.
166.
by Leynos 11mo ago
Having larger models is nice because they have a much wider sphere of knowledge to draw on. Not in the sense of using them as encyclopedias. More in the sense that I want a model that is going to be able to cross reference from multiple dom
167.
by Leynos 1y ago
I think that you have to be OpenAI (or X, Google or Anthropic) to be able to fine tune models of this scale through reinforcement learning at present. Look at Tinker for an example of where things might be heading though ( https:/&#x2F
168.
by Leynos 1y ago
That's pretty much how I use Codex.
169.
by Leynos 1y ago
That should read "Turing created the Bombe" https://en.wikipedia.org/wiki/Bombe
170.
by Leynos 1y ago
I wonder if a large context model could be employed here via tool call. One of the great things Gemini chat can do is ingest a whole GitHub repo. Perhaps "before implementing a new utility or helper function, ask the not-invented-here
171.
by Leynos 1y ago
What do you think was the reason that the bootcamp grads struggling to get better at what they do?
172.
by Leynos 1y ago
The allegation is that the party in question copied the author's code verbatim and stripped off the copyright / licence notices.
173.
by Leynos 1y ago
MIT places the following condition on the licencee if they wish to re-distribute the code: > The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. Which the other p
174.
by Leynos 1y ago
I mean, the AI companies have £200 a month plans for a reason. And if you look at Blitzy for example, their plans sit at the £1000 a month mark.
175.
by Leynos 1y ago
It depends what you're fitting to. At the simplest, you can ask for a reduction in cyclomatic/cognitive complexity measured using a linter, extraction of methods (where a paragraph of code serves no purpose other than to populate
176.
by Leynos 1y ago
When an agent can work independently over an 8 hour day, incorporating new information and balancing multiple conflicting goals—then apply everything it learned in context to start the next day with the benefit of that learning, repeat day
177.
by Leynos 1y ago
Apologies for the second reply, but it also occurs to me that reinforcement learning is the new battleground. Look at the changes between o1, o3 and GPT-5 thinking. Sonnet 3.7, Sonnet 4, and Sonnet 4.5. And so forth. I expect models will ge
178.
by Leynos 1y ago
I suspect continuous learning will be the next driver of GPU usage.
179.
by Leynos 1y ago
Don't YOLO unless you're comfortable letting a manic depressive chaos goblin loose on your system :P
180.
by Leynos 1y ago
Seems like they're shooting for feature parity with Claude Code, but open source and with a hooks-based plugin system

More ›