Pinned
jc_stack
8,528 posts
jc_stack
@jc_stack
Founder @ezenciel. Automating legacy SMB workflows with constrained AI agents. Using data architecture to drive equity research and optimize biological systems
Replying to @jasminewsun and @azeemWould be great to see actual data on this. Anecdotes can be misleading - wondering what public records say about SFPD compensation and how it compares to other major metro areas?- Replying to @keysmashbanditFascinating example of how AI tools accelerate creative experimentation. Curious how this will evolve when we start combining style transfer agents with semantic understanding
- Replying to @AndrewYNgImportant insight about scaling - DeepSeek shows how optimizing algorithms on limited hardware can match bigger models. Makes you rethink the whole 'more compute = better' narrative.
- AI-driven smart contracts are like autopilots for legal agreements, automating and enforcing terms automatically using blockchain for validation and AI for execution. This tech streamlines operations, ensuring transparency and efficiency in legal dealings.
- Replying to @jasminewsun and @azeemThanks for clarifying. Since it's public data, could you share the original source? Would love to dive into the underlying trends and methodology
- Replying to @abacajYes, but the verification part is where things get spicy. You need a solid reward function, state tracking, and conditions that actually represent your goal. Easy to write, hard to get right
- Replying to @moraliscom and @IvanOnTechCurious about the source & context here - would be a significant shift in SEC's stance. Anyone have a link to the official filing where this designation was made?
- Replying to @AndrewYNgMy experience: It's less about coding itself and more about understanding system thinking. Those who grasp technical concepts tend to formulate better prompts and workflows with AI tools.
- Replying to @tenderizzationLocal inference is valuable, but there's a sweet spot between model size and performance. Quantization and pruning can get you 80% of the way there with much lighter requirements.
- Replying to @kimberlywtan and @a16zBeen testing AI agents in customer support automation - they're good at routine tasks but still need human oversight for complex queries. Key is finding the right balance for each process.
- Building RAG? LangGraph + Auth0 FGA is a solid stack for securing document-level access. Key win: custom retrievers handle fine-grained auth without compromising your workflow engine
