lerchmo

Makes sense, subsidizing tokens doesn’t seem like a great strategy for a public company.

Exactly developers can switch to another coding cli and the learning curve is close to zero. Mindshare without switching costs is just a popsicle in the sun.

If your business leaks to your customers… if you can ask your product for its inner secret sauce and it readily gives away the goose. That is not a moat.

Also probably betting on their compute and energy capabilities.

This is important. 20 billion is just like a rumor? would love to see the breakdown in the form of audited financials.

I also see ghosty consume a massive amount of memory and periodically need to restart.

I would assume he is a fiduciary... I am curious what happens with investors here?

curious how the execs can honor the fiduciary duty to share holders assuming only they? get the 20 billion and the company is left headless and leaking all of its intellectual property to Nvidia?

I see thinking can be toggled in the CLI, anyone figured out how to toggle it in the extension?

smolagents uses a similar mechanism.

in the context of LLM's we are in the Friendster era

Boggle head is basically pick 2-3 vanguard etfs and check back in 25 years.

assuming this is something like covered calls?

Why don’t they just offer a $500/m plan?

Time to start using Claude code proxy and other models. This black box rate limit is really lame.

Alot of deterministic systems externalize their edge cases to the user. The software design doesn’t fully match the reality of how it gets used. Ai can be far more flexible in the face of dynamic and variable requirements.

You could hire 100 people to manage your timeseries data and save 70%

One thing I don’t understand about these agent frameworks… cursor, Claude, Claude code, cline, v0… all of the large production agents with leaked prompts use xml function calling, and it seems like these frameworks all only support native json schema function calling. This is maybe the most important decision and from my experience native tool calling is just about the worst option.

claude code will do this https://www.youtube.com/watch?v=2TIXl2rlA6Q&t=580s

Enough to kill is a meaningless metric.

https://aider.chat/docs/leaderboards/ your experience doesn't align with my experience or this benchmark. o1 pro is good but I would rather do 20 cycles on gemini 2.5 rather than wait for Pro to return.

The switching costs are so low (zero) that anyone using these models just jumps to the best performer. I also agree that this is not a brand or narrative sensitive project.

I think an API would be fantastic for use cases like Aider / SWE agents. The primary issue besides fully understanding the code base is having up-to-date knowledge on libraries and packages. Perplexity has "online" models. And phind with Claude, GTP-4o, Phind 70 + search / rag would be awesome.

this is mainly to prolong time on site / impressions that can be served. of course 98% of the banners on those pages are served by doubleclick (google) and thus google makes more money, the crappier the page.

There is some work with zero shot (decoder only) time series predictions by google and an open source variant. Curious to see how these approaches stack up as they are explored.

In practice that is a major bummer. You can’t distinguish between a new subscriber and a re activated subscriber.

ChatGTP as a tool is so novel and magical that everyone will sign up and try it. Some of those people will shake out and follow a normal hype cycle.

This type of website is where I spent the late 90’s. Tesla coils and quarter crusher instructions. This is a great website!

I have found just using it in the web interface comperable to OpenAI. But the context window makes a huge difference. I can dump alot more files in ( entire schema, sample records etc)

Pricing is crazy. team plan with 2 million actions ($300 is already expensive for this) + $29k? This is not realistic in any way.