@levie One challenge is that frontier model companies are not incentivized to make their harnesses token efficient. But thats avenue for other companies to step in and do it with open source models once they catchup. reason to be bullish for xAI/Cursor.
@magicsilicon I think what could make NVIDIA's moat more fragile than Intel's is that CPUs dominated majority of computing for decades. We don't know the final nature of AI computing. Training is giving way to inference to agentic computing with CPUs/memory hierarchy coming back to focus.
@dharmesh love this!! But also cannot help thinking this is probably a very low risk decision for you, given your overall wealth. The real high risk decision you have already made is to continue leading HubSpot even after 20 years.
Today, we're launching Incognito Chat with Meta AI on @WhatsApp and the Meta AI app, a new way to have completely private conversations with AI.
https://t.co/VZWbAx60th
@Chillestdotcom@Nick_Davidov true and you would always know the total fee/carry across layers upfront. You dont have to do your own math to figure it out. I think people are making a bigger deal than they should with this.
@antoniogm I can understand being stuck in traffic but what is a needless loop? I think it is great feature to cut the ride short and pay for the portion you took + an additional fee :) humans get pissed but AVs should not.
@pitdesi I feel like this is a great way for governmental and non-profit organizations to get funding (a lot less controversial than taxes), everyone is happy. Why doesn't this happen more often?
@jasonlk@jerseejess@Work_Bench I a human needs to decide whether to use your new software or not it is extremely hard to win. I think the future is to convince the agents to use your software and recommend it. It is a different GTM strategy.
@philippehebrard totally agree. and I keep reminding myself to tell the agent remember this for all future interactions and sessions. It doesn't always work. Also I feel like the agents should be smart enough to know what is temporary vs. permanent or I am expecting too much :)
Is it me or is Claude code getting worse? I kept getting frustrated by CC forgetting context and making wrong decision on a personal project. switched to Codex a few days ago. Significantly better! Have not had a single issue yet. I am not sure it is the model or the harness!