Pinned Tweet
My thoughts on DeepSeek's impact to the market and scientific community:
BLUF: Short term de-risked into uncertainty, long term hugely bullish for AI and will cause mindshare to swing fully back from meme to AI.
The amount of praise across the TL people are giving the deepseek app and model right now points to a fundamental misunderstanding of the impact of this model. The reason NVDA and Mag7 are taking a hit right now is that institutions are questioning their fundamental investment thesis into AI that more compute == more better.
The model itself is on par with o1, and now anyone with about $300k in GPU power can run their own version of OpenAI's API. Crazy. Huge. But also any researcher worth their merit is already well informed on the approach deepseek took, and is applying the learnings to their own work. Soviets got Sputnik into space first but the US was the first on the moon.
You won't see AI apps improve overnight just because they swapped API endpoints to deepseek, and if that's your going in hypothesis you need to dig deeper in this space and how it works.
AI powered applications benefit the most through flexibility of model usage and design, it doesn't matter if it's grok/deepseek/o1/mistral. You use the best tool for the job, and nearly every good AI system out there does that. Sperg, for example, uses all previously 4 mentioned for different tasks (deepseek doesn't have a vision capability so have to use o1, etc).
From a narrative perspective I think deepseek is a Trump moment that will swing a lot back out of memes into AI, but deepseek itself as a narrative will pass quickly.
A few updates on $Sperg:
For a while, Wu has been using indicators as part of his toolset for looking at charts. Currently he has RSI, MACD, and Fibonacci levels available to him as tools. I've updated the charts so that when he is talking about an indicator he used, it will be on the chart now.
Congrats @zoopee1 and @ddupont808 , it has been awesome and inspiring to see what you all have been able to do with this project. Looking forward to seeing your LAMs take over the web.
WOW!
$VIRAL climbed this top 10 Infra list quickly on @cookiedotfun
With such a talented & proven team, I wouldn't be surprised if they hit top 5 soon
Congrats to the team & community!
Just for clarification, as I realize its been a couple days since I gave an update, the poetry feature @we_toddit has been having fun with is not what I've been working on, he's just having fun.
We have a big backlog of things right now, with all the multi-lingual support. Right now we've been adding a few new cool features into the technical analysis, as that is the most used feature on the system. Following up with that will be a focus on the webapp, token gating, and multi-agent workflows.
Okay remember how I said things might get messy?
First, floored by the overwhelming support here. Big thank you to @AethirCloud for having us in their Ecosystem Fund, really excited about where that will go.
Second, in true $SPERG fashion we released major updates right before a big influx of usage. A few issues have popped up that I'm addressing right now:
- The translation layer is having issues working on X, (it works in TG),this should be fixed shortly so all the bots can reply in the native language of the query
- There are some limits on how many times the agents can respond (to avoid infinite agent loops), this may be getting in the way of some responses, digging into this.
- @SpergReporter appears to be flagged as spam, likely due to low follower count (give him a follow please!). Working on paths forward there but should be resolved when he reaches enough follows
Working on all that now. Hope you're all having fun. I am.
This is a fairly different way of multi-agent interaction than what people are use to. Tagging our swarm manager will route to (one for now) agent to answer.
The exciting part about this is how easy it is to add fully orthogonal functionality into the same system with the same UX (What if we want degen art generation too?).
I'm sure there will mistakes and messiness along the way, but should be fun.
This quoted post is unavailable.
We have so many individual agents, out there today that have comparable functionality and features, many that are very strong (re: agents like @aixbt). I think the real undiscovered value is how you use them together to gain an edge, and the frameworks that help you do that.
Some details on what I've been doing behind this.
Right now we have two agents public (a third unreleased but part of the workflow used here)
I've been developing the core manager system to create specific workflows for data gathering. In particular the one that feeds the content for these videos. You might hear terms like "swarm" or "MAO (multi-agent orchestrator), this is the system that I've been building out to allow for user-defined arbitrary workflows to be created that utilize the agent's tools.
In this one, I prompt the manager to generate a news report with at least 5 recent crypto news items, it then coordinates the agents to search , gather, and aggregate both content and images for each. Which I then feed to the avatar generation video system that ultimately creates this video.
Our website will soon be updated and token-gated, and I'll be releasing the workflow prompting as a beta to a number of $SPERG holders for feedback and iteration.
This quoted post is unavailable.