Marcin Krzyzanowski · pinion

the fact that http://pi.dev agent is so good, with virtually no sophisticated harness whatsoever, is a testament to the fact token vendor (codex/claude) agents are overrated. highly. today's moat of codex/claude tools is GIVING TOKENS FOR FREE LEFT AND RIGHT to make you dependant

Marcin Krzyzanowski@krzyzanowskim

the fact that pi.dev agent is so good, with virtually no sophisticated harness whatsoever, is a testament to the fact token vendor (codex/claude) agents are overrated. highly. today's moat of codex/claude tools is GIVING TOKENS FOR FREE LEFT AND RIGHT to make you dependant

related

▶@coderabbitai

@om_patel5 I've tried all ( 74 😵‍💫 ) AI Coding Agents & IDEs [Rork, CodeRabbit, Anima, Zed, Factory, Cursor, Windsurf, Copilot, Lovable, Bolt, v0, Replit, MarsX, Canva, Devin, Github Spark, Vercel, Lindy, Warp, Figma, Cline, Vibe Coder & more] The most complete list ever made (with demos & notes):@johnrushx

▶@whosjunaidd

▶@coreyhainesco

▶@TheAhmadOsman Techniques I'd master if building RAG systems that actually work: Bookmark this. 1. Sliding Window Chunking 2. Semantic Chunking 3. Document Hierarchies 4. Metadata Enrichment 5. Query Expansion 6. Hybrid Search 7. Reranking Models 8. Context Window Packing 9. Lost in the Middle Problem 10. Hypothetical Document Embeddings (HyDE) 11. Multi-Query Retrieval 12. Contextual Compression 13. Sentence Window Retrieval 14. Auto-Merging Retrieval 15. Cross-Encoder Rescoring 16. Temporal Context Decay 17. Negative Sampling 18. MMR (Maximal Marginal Relevance) 19. Graph-Based Retrieval 20. Recursive Retrieval 21. Citation Trackingchunks 22. Context Ablation Testing 23. Adaptive Retrieval@athleticKoder

▶@abouelatta_ali

▶@thdxr Top 5 local LLMs: 1. GLM-4.5-air: best agentic/coding model that runs on consumer hardware at very decent speeds. Rivals Claude 4-sonnet. 2. Nousresearch/hermes-70B: the only model that will do whatever you ask, and tell you whatever you want to know. Literally critical to have. 3. GPT-OSS-120B: very intelligent it’s like having 4o at home, great context window, great agent 4. Qwen3-coder-30B-3A-instruct: very good coding agent, excellent workhorse, incredibly fast 5. Mistral-magistral-small: very fast, excellent agent, great coder, multimodal, punches way, way, above its size. I would be okay never using a proprietary llm, although given the subsidised compute I will continue to use them since I’m getting free leverage.@0xSero

@bbssppllvv I took the @karpathy autoresearch loop and pointed it at markets. 25 AI agents debate macro, rates, commodities, sectors, and single stocks daily. Every recommendation scored against real outcomes. Worst agent by rolling Sharpe gets its prompt rewritten by the system. Keep or revert. Same loop, prompts are the weights, Sharpe is the loss function. Trained the agents on 18 months of market data. 378 iterations. 54 prompt modifications, 16 survived. The system learned which agents to trust using Darwinian weights — geopolitical, commodities, and the @BillAckman quality compounder rose to the top. The agents even figured out their own portfolio manager was the weakest link before we did! Deployed the trained agents. +22% in 173 days. Best pick: AVGO at $152, held for +128%. The final prompts are evolutionary products — shaped by market feedback, not human intuition. Now running live with my own capital. https://github.com/chrisworsey55/atlas-gic Part hedge fund, part research experiment :)@Chris_Worsey

▶@askalphaxiv

@realmcore_@uAghazadae @Zai_org should work perfectly with http://eve.dev / http://ai-sdk.dev@rauchg Vercel Ship (so far) ▲ Vercel Services — Run full stack on Vercel — Microservices on Vercel — Private service comms ▲ eve — The framework for building agents — Durable agents with one folder — Sandboxing, approvals and evals ▲ WebSocket support — Persistent connections ▲ Vercel Functions — Dockerfile support — 5 GB packages, 30 min — Zero-config Node servers ▲ Vercel Agent in Dashboard — Autonomous investigations — Plan-based permissions — Fixes surfaced as PRs ▲ Vercel Container Registry — Container images on Vercel — OCI push, pull and tag ▲ Vercel Sandbox — Custom images — Persistent sandbox storage — 24 hour runtime + expiration ▲ AI SDK 7 + agent harnesses — Toolkit for building agents — Coding-agent harnesses ▲ Chat SDK — Multi-channel agents ▲ Access & security — Vercel Passport: OIDC access — Vercel Connect: external services — OIDC audiences + Audit Logs ▲ Enterprise apps and agents — Security posture dashboard — Vercel Functions in your cloud ▲ AI Gateway — Voice, speech and transcription — API key budgets ▲ Vercel Blob — Private blob storage — Signed URLs and OIDC auth ▲ Builds & deployments — 500 concurrent builds for Pro — Vercel Drop — Elastic builds ▲ Vercel CLI — Speed Insights + Web Analytics — Trace any request — Deployment limits removed ▲ Observability — Trace eve sessions — Trace Workflow runs ▲ Workflows — Inflight cancellation — Nitro v3 + TanStack Start ▲ SDK-free Vercel Flags — No SDK keys@ctatedev @wesbos definitely @mobbin https://mobbin.com/ also UI library showcases like @chakra_ui is an underrated place to look!@jeffzxh I built http://markdown.new Put http://markdown.new/ before any URL → get clean Markdown back. Cloudflare's Markdown for Agents is great, but only works for enabled sites. http://markdown.new works for ANY website on the internet. 80% fewer tokens. Also converts PDFs, images, audio. Free. No signup. http://markdown.new@elbeyoglu