Dr. CryvΞ
@drcryve
Followers
1K
Following
2K
Statuses
1K
Dr. | AI Strategist & Agent Expert | Token Engineering Lead @OVioHQ | Views are my own, not legal or investment advice
Joined December 2021
🪙 How can we match the token vesting with the actual demand for a token? 👀 Explore it in my latest piece on Adoption Adjusted Vesting (AAV) 💡 🫳 #tokenengineering #crypto
@OVioHQ @tokengineering @tokenomicsdao
1/5🚀 What are the right vesting terms to satisfy early investors and to maintain a healthy token economy? A new Adoption Adjusted Vesting (AAV) approach could be the game-changer you need. Read the TL;DR of @drcryve's latest article 🧵📝
1
4
17
RT @DoggfatherCrew: @drcryve Amazing! I will def dive into your talk! Love your (and your company's) thesis and need to find all the alfa i…
0
1
0
@DoggfatherCrew It was yesterday and they recorded the whole conference. It will be online soon, I guess. Rich Sutton, the founding father of reinforcement learning, was speaking there as well, along with many other industry legends. 🔥
1
0
1
@hananyss_ @virtuals_io I am happy for you seeing you are embracing your next adventure! Every business that is working with you is privileged.💯
1
0
0
RT @phtevenstrong: Why I'm so bullish on Christianity: First, I compared secular / non-secular market data. Virtually all major indicato…
0
143
0
Great to see the positive feedback on my approach to building a convenient, simple, and flexible solution for multi-agent systems with PydanticAI 🤩 👉 If you’d like to check out the minimal example mentioned, take a look at the repo: Also, if you're working on a business in the multi-agent system and DLT space, I’d love to connect—feel free to comment or DM me! #AI #DLT #AIAgents #MultiAgents #AGI #DeepSeek #PydanticAi #chatGPT #openAi
0
1
2
RT @OVioHQ: 🔔 Reminder: Join us TODAY for @Cryptic_Web3 🎙️ Don’t miss this deep dive into the future of AI & Web3, with @drcryve now speak…
0
2
0
RT @0xAkina_: AI Agent in Gaming: The Ultimate Overview 👇 INFRA/FRAMEWORK • ArcAgent (@ARCAgents): An open-source innovation movement pow…
0
6
0
AI is now commoditizing itself—an inevitable milestone on the road to AGI. DeepSeek demonstrates that distributed LLMs, composed of a specialized sub-module design, are more efficient than monolithic models such as ChatGPT. In its early stages, AGI will likely emerge as a system of expert agents consisting of modular LLMs, each dedicated to a specific domain or even granular task. As the cost of LLMs decreases and accessibility grows, limitless specialization becomes possible, driving unprecedented performance across all fields. #DeepSeek #AI #AIAgents #OpenAI #chatGPT
0
0
1
RT @OVioHQ: 5/7 @drcryve, our Token Engineer Lead, emphasizes the economic impact: “@deepseek_ai competes with leading models while requir…
0
1
0
RT @OVioHQ: All you need to know about @deepseek_ai from @OVioHQ in-house hive mind. Why does it matter? It’s not just another AI model —…
0
4
0
DeepSeek-R1: Why It Falls Short for AI Agents and How We Can Fix It 🤔 DeepSeek-R1 has generated significant buzz for its impressive performance and affordability. But there’s a catch—its lack of tool support makes it unsuitable for advanced AI agent systems. Here's the issue and how the open-source community is stepping up. 🛠️ ♻️ Quick Recap The release of DeepSeek-R1 is revolutionary, with the widely cited $5.6M figure for its latest training run. However, it’s important to note this excludes critical costs like GPU acquisition, earlier model iterations, and infrastructure. Even so, DeepSeek-R1 competes with industry leaders like OpenAI, Meta, and Anthropic on key benchmarks such as MATH500 or CODEFORCES, all while keeping costs significantly lower. Its open-source nature further accelerates innovation across the community. 🛠️ The Issue: No Tool Support While DeepSeek-R1 excels at basic workflows and chat-based systems, it lacks tool support. This is a major limitation for AI agent systems that rely on tools to perform tasks like database queries, API calls, or advanced custom logic. Without tool integration, its potential for real-world, high-impact applications is restricted. ☺️ The Solution: Community to the Rescue Fortunately, DeepSeek-R1 is open source, and the developer community is already addressing this gap. For instance, models like "michaelneale/deepseek-r1-goose" (available on Ollama) introduce tool capabilities. You can find this and similar models with tool support through this search filter: In my experience this approach has delivered the best results for tool utilization for DeepSeek-R1, making it a great option for those looking to integrate tools into self-hosted or local systems. Alternatively, for those needing in-built tool capabilities by default, you might consider DeepSeek-V3. However, for advanced use cases, models with >14B parameters are usually required, which means most regular laptops won’t cut it. 💰 When Budget Is Not a Concern For AI agent systems where budget isn't a constraint, my top-performing choice remains the anthropic/claude-3.5-sonnet model. It delivers exceptional results but comes at a significant cost. If you're a developer or founder building cutting-edge AI agent systems, let’s connect! We help early-stage teams get up to speed with a strong network of experts and investors. Learn more here: #DeepSeek #DeepSeekR1 #AI #AIAgents #OpenSourceAI #AgentSystems #DLT #StartupInnovation
1
0
1
A new libertarian world order establishes in front of our eyes. I hope that it'll bring the promised freedom and, consequently, prosperity while still setting boundaries where necessary. Most importantly, I pray that all conflicts may be solved via diplomatic or economic measures instead of military ones. 🙏
0
1
3
0
0
1
0
5
0