I find that it consistently breaks around that exact range you specified. In the sense that reliability falls off a cliff, even though I've used it successfully close to the 1M token limit.

At 500k+ I will define a task and it will suddenly panic and go back to a previous task that we just fully completed.

The context7 MCP helps with this but I agree.

Interesting that you're migrating assistants and threads to the responses API, I presumed you were killing them off.

I started my MVP product with assistants and migrated to responses pretty easily. I handle a few more things myself but other than that it's not really been difficult.

I like hotels but I love renting someplace unique and pretending that I live in that city or place for a week or two.

I understand your point if all you're looking for is somewhere to sleep that's clean and comfy.

The comparison table on the kilo site says "OpenRouter without 5% markup" and only puts a checkbox next to kilo.

I can't find any reference to Cline/Roo charging anything on top of API pricing.

Not sure how they'd do it considering you bring your own API keys. Can you link me to a resource?

Even with proper tool call descriptions, I've had quite a few occasions where the LLM didn't know how to use the tool.

The tools provided by the MCP server were definitely in context and there were only two or three servers with a small amount of tools enabled.

It feels too model dependant at the moment, this was Gemini 2.5 Pro which is normally state of the art but has lots of quirks for tool use it seems.

Agreed on hoping models are going to be trained to be better at using MCP.

We don't know whether pushing towards AGI is marching towards a dystopia.

If it's winner takes all for the first company/nation to have AGI (presuming we can control it), then slowing down progress of any kind with regulation is a risk.

I don't think there's a good enough analogy to be made, like your nuclear power/weapons example.

The hypothetical benefits of an aligned AGI outweigh those of any other technology by orders of magnitude.

I think I linked to the verification process on OpenAI's website in then documentation.

You need to verify your identity with a driver's license or passport with OpenAI to have access to certain things like chain of thought summaries in the API and image generation with the new model.

Nothing I can do there, you gotta verify unfortunately.

Taken to the extreme and dumbed down, I don't think it's bad to think about it like that to be fair. HYPE spewing aside.

The comment you replied to is sarcastic but magic box that does everything is pretty much where things end up, given enough time.

I built a local playground for it if anyone is interested (your openai org needs to be verified btw..)

https://github.com/Alasano/gpt-image-1-playground

Openai's Playground doesn't expose all the API options.

Mine covers all options, has built in mask creation and cost tracking as well.

Not sure to be honest, but it feels like if HN never goes through a redesign, we'll still be enjoying it in 2050.

I think my positive perception of self named software dates back to Irfanview.

I've been using it since 1998, never felt the need to change to anything else.

The wiki entry is pretty short and sweet on purpose and potential flaws.

https://en.m.wikipedia.org/wiki/Military_wargaming

I'm sure that there are many amazing products on this list but I just want to give a shout out to Typesense (Algolia alternative) from personal experience.

The product is great and the founders are incredibly attentive to their community.