@ThePrimeagen false comparison + 2 things can be true. AI likely doesn't 100x most people, but also AI isn't used only for decision making, it's mostly used for grunt work where a 100x improvement is 100% possible.
@rucam365 Claude code is still my/our goto for coding. Sonnet handles most tasks, Opus as needed. Heard Codex is now better but OpenAI upset me, plus I mostly use what my customers provide and enterprises move slowly.
@AMANPRE37638190@BenBajarin As a software guy I'm predicting the AI trend, no idea how to invest it yet.
HBM and VRAM are key with AI regardless of datacenter vs edge. I did buy EMXC to get more direct exposure to non-US semis such as SK Hynix and Samsung
@davepl1968 A lot of people acted badly during that time. It was as much a political pandemic as a health pandemic so that gave reason for irrational people to be even more irrational than normal.
My .02c. Forgive, learn, move on :)
Prediction: 5 years from now Devs will run most (80-95%) of their AI workloads locally.
Today's OSS coding models are already good enough for most problems. They will be far better in 5 years.
Today's hardware can already run coding models. All you need is a decent GPU, a DGX sparc, a Jetson Orin Nano Super, etc. Hardware will be optimized for AI workloads and performance for AI will increase far faster than Moore's law.
Today we pay for Claude to get the harness (Claude code), but that piece will have strong OSS alternatives. This is now especially easier since much of the harness can be generated using AI.
You will be able to tap into datacenter level compute/memory or frontier models as needed but will seldom need it for coding.
If I'm right, there are some interesting questions.
1. As an investor, I'd bet on OpenAI, Claude, others. for the next 2 years, but for 5 years...it's a coin flip. At least one ends up as myspace(dies), one ends up as Facebook(prospers), and one ends up like reddit (survives but meaningless). I guarantee a big player like Oracle, Google, Microsoft, IBM buys at least one of them.
2. What do we do with all the compute we are building out now? My guess is they become compute wholesalers to corporations that have AI baked into their apps and the AI datacenters become a new Cloud, call it Cloud 2.0.
@lovelacecoding you did miss out; they were simply amazing. In many ways even a decade later, the best iPhone and android still haven't caught up. It was a masterclass in design, and also a masterclass in how not to execute.
@MrPeterLMorris@sadhna6389 respectfully disagree. The level of financial expertise that $3K/year buys you from an advisor is practically zero. They take their fees then put you in a TDF, balanced 60/40, etc.
@mcuban Wish we had you as a healthcare affordability czar but then realized you would be expected to divest. Somehow, we have to find a way to get smart/capable people like you into roles where you influence policy.
@JeffBezos Failure happens. It is a part of life.
Winners embrace it, learn from it, grow stronger from it, and eventually start to seek it out when they realize if you aren't failing at least occasionally, you aren't pushing yourself hard enough.