@TheAhmadOsman Please Tell us what do I really need to service 5-12 people at minimum?
English
sven2401
3.5K posts
No, you won’t be able to “service 5-12” people with a bunch MacBooks hooked to each other
That larper didn’t even know what an MoE is 4 months ago and now he’s acting like an expert and spreading misinformation
When did lying become so normalized in this space ffs
English
@Sentdex Correct but if the anthropic employees say that it’s fine and then you put effort in and get rug pulled that’s frustrating. It wouldn’t be so bad if it would be like that from the beginning. If they cleared all that up in January after they banned opencode usage
English
Apparently an unpopular opinion, but I don't think Anthropic owes anyone heavily subsidized tokens for their third party app.
Theo - t3.gg@theo
I can't help but feel personally burned by the Claude Code changes announced today. We put so much work into wrapping the (atrocious) Claude Agent SDK in T3 Code. It was the ONLY path they supported, so we made it work. It was hell. Now our users are getting their rate limits cut by 40x, despite us doing everything right. I listened to the Claude Code team. I had my issues with their direction, but I trusted them and took them at their word. I will never make that mistake again. Until we see significant change, it is safe to assume any statement from an Anthropic employee is a lie on a timer. The rug will be pulled, no matter how many promises are made beforehand.
English
@lucasmeijer Not sure how implements slash commands, but skills invoked with a command have the additional feature that they are not only a prompt file but can have additional support files in that skill folder the agent can discover if needed.
English
i never use skills. almost everything is a pi slash command instead.
English
OpenCode Go will do 1.5T tokens today
Our free users will do close to 2T tokens
Pay-per-token users are close to 500B
Also, this is just a part of our user base because we don't track your usage if you don't use our inference
Jay@jayair
OpenCode Go is now our second breakout product Closing in on 1T tokens per day Shoutout to Frank and the team
English
@KingBreton78876 @thdxr Likely a combination of model and manual compaction?
English
@thdxr What do you mean "Compaction is good now"????
Where's the news, that OpenCode has a compaction that does not cause lobotomy??
English
@bukitsorrento @Howaboua The requests is only an estimate it’s not request based, afaik. You run through „dollar“ too.
English
@mattpocockuk I added the specific call to it in the prompt for use with GitHub copilot. I can make a grilling session with gpt 5.4 xhigh for just 1 request with the cost equivalent of couple of cents. Need to squeeze as much as possible until end of month.
English
@LottoLabs With API cost one small task can already cost you 1$ as it wastes thinking tokens like crazy. 4h work for 6$ seems cheap in comparison
English
Opencode Go update $10 subscription
Usage limits are definitely lower than ollama cloud and that makes sense for half the cost but I dug (slightly) deeper tonight
I did 4 hours of work today with k2.6 and billed $6.59 that’s out of $30 weekly limit and $60 monthly limit
One issue is finding the pricing of the model $/M tokens not listed, and it isn’t API costs. You can see examples of input tokens, output tokens in the screenshot and they average around $35-50/M tokens, this is obviously way above the $3.5-4 for api.
At $35-50/m tokens and the limits in place you can get roughly 1-2M tokens out in a month. If you just used API that would cost $3.5-7 a month anyways. Obviously you get access to other models and it’s hosted fairly by opencode etc. but it’s not as crazy a deal as people make it out to be.
I still recommend it, especially the first month because you actually are net ahead at $5 minus getting rate limited.
Unless I’m missing something with input output costs you basically get $10 equivalent in api anyways.
English
@0xSero helped me setup local models properly and I uh, had no idea these things had gotten this good
Are they frontier level? No, but considering this is running on just my 5090 it's remarkably capable
First tests on a couple of programming tasks and the qwen 3.6-27b model with no reasoning feels about on par with something like sonnet 4-ish, probably better it's really impressive
But also setting up local models isn't easy, I don't know nearly enough to talk much about it yet other than you need to know what you're doing to have a good experience. The out of the box stuff is not nearly as good as setting it up correctly
English
@mattpocockuk And what if you don’t have a ui in the general sense (not programmable at least)? 🤔
English
The more I replace plans with prototypes, the better the outputs
Who'd have thought that low fidelity prototypes were better than walls of spec
Oh yeah, the entire industry for 20 years
Stop going against decades of knowledge because someone in SF shipped it as a 'mode'
dax@thdxr
i never make plans i hate looking at markdown i don't wanna read markdown files i just plan by having it make changes to the code then i look at the code to see what sucks then i prompt again
English
Mini-realization this morning
Whenever you're QA-ing code produced by an AFK agent, you're not just QA-ing the code...
...you're also QA-ing the agent itself. So, fixes must land in both in the same commit.
English
@mattpocockuk Maybe yes as a free course it might even boost your other courses.
English
Sounds mad, but maybe I should just make a course about writing great skills?
I.e. for actual life/work productivity, not just dev.
Breaking down daily tasks into skills. Turning HITL tasks into AFK ones. Creating a working language with the agent.
Feels pretty deep
English
@MrAhmadAwais But that was different. Kimi K2.5 was used hugely by them after its release in January
English
Seems like the OpenCode team doesn’t use open models much themselves. 😅
dax@thdxr
our teams last 7 days of spend damn gpt5.5
English
@alikonkret @2000WU6 @sparbuchfeinde Ich weiß nicht wie das in dem Fall ist aber Arbeitszeit und Arbeitszeit. Reisezeit ist nur dann Arbeitszeit für das Arbeitszeitgesetz wenn du hinterm Steuer sitzt. Trotzdem wird man normalerweise für die ganze Zeit bezahlt.
Deutsch
@2000WU6 @sparbuchfeinde "Ausbeutung" ist mir etwas zu hoch aufgehangen. Es ist halt so in der Leistungsgesellschaft. Erst recht, wenn man ergebnis- und nicht nur verrichtungsverantwortlich ist.
Deutsch
4h zum Kunden fahren, 4h Meeting und Business-Lunch, 4h wieder nach Hause fahren.
In jedem Land der Welt ein anstrengender, aber normaler Arbeitstag.
In Deutschland ein Verstoß gegen das Arbeitszeitgesetz.
Und du bist als Vorgesetzter oder Mitarbeiter direkt in der Mithaftung, wenn etwas passiert.
Es nervt einfach nur. Und muss dringend reformiert werden.
Wenn Gewerkschaften wirklich etwas an ihren Mitgliedern gelegen wäre, würden sie wegen der absurd hohen Steuerbelastung oder der Rente auf die Straße gehen.
Deutsch
@championswimmer The thing is at what cost. Kimi k2.6 throws tokens out of the window like it would be nothing. Compared to K2.5 don’t like that. Cost and time to finish manyfolded.
English
Cursor Composer 2 is fine-tuned Kimi K2.5
Kimi K2.6 can also be basically defined as the same thing? (Essentially sharing the same pre-trained base model lineage + new post training)
Cursor hosts the model on Fireworks.
So if you use Kimi K2.6 directly on Fireworks yourself, then you are sorted. What is the Cursor moat? Just that it has sold subscriptions and has some captive audience? And benchmarks below is a lesson that it is hard/impossible to beat the model maker themselves at fine tuning their own model (unsurprising).
English
@r_marked Why does the EU allow refunds under a 14 day period lmao what's the reasoning behind it
English
- 15 messages
- $221 of tokens
- 1.6% of my $40 plan used
It's obvious that GitHub couldn't keep this model for billing on Copilot.
Theo - t3.gg@theo
I sent a single message on Copilot and it did over 60m tokens. It's still going. $30 of inference so far. In their current billing model, you get 1,500 messages, regardless of how expensive each is. I'm pretty sure I can do $45,000 of messaging on this plan
English
@mkurman88 @burkeholland That is so true. I can basically pay per token on openrouter at OpenAI them selfs or anywhere. The only benefit is to not have to have all the api keys but that openrouter already solves. And yeah you get free autocomplete but that’s not somethings someone is staying for
English
@burkeholland If you somehow make it more affordable, I will stay. For now, the planned prices are as high as OpenAI's API. Why should we stick to your subscription if it doesn't give us any profits yet restricts us with an AI token limit that doesn't even accumulate?
English
After the copilot sub becomes useless, I’m fully switched to DeepSeek.
English