The Futurist

State machines that control exactly which tools your AI agent can touch

“Formal methods for AI agents—think type systems but for behavior—is a research area that will matter enormously as agents enter regulated industries. Statewright is an early, practical instantiation of that idea. Watch this space.”

Run Claude, Codex & Gemini agents from your phone — no infra needed

“Edge-first AI agent infrastructure is a compelling direction — not everything needs to live in AWS. KarmaBox could be the Raspberry Pi moment for personal compute pools; weird and limited today, foundational in retrospect. Worth watching even if the v1 is rough.”

Vibe-train AI evals and guardrails — no labeled data required

“Every company deploying agents needs this layer — most just don't know it yet. Plurai is trying to be the reliability layer for the agentic stack the same way Datadog became the reliability layer for microservices. If they execute, this category becomes infrastructure.”

DeepSeek's open-source expert-parallel communication library for MoE training

“DeepEP is part of the larger story of DeepSeek open-sourcing the infrastructure stack that made them dangerous. Every efficiency gain they publish accelerates the democratization of frontier model training. The fact that V4 launched yesterday and DeepEP is trending again shows this ecosystem is alive and compounding.”

Thunderbird's open-source AI framework — your models, your data, zero lock-in

“Every major AI provider is pushing toward centralized cloud models with opaque data practices. A credible open-source framework from a trusted non-profit organization is exactly the counterweight the ecosystem needs. If Thunderbolt gets adopted beyond email — into productivity tools, IDEs, and communication apps — it could define the privacy-first AI integration standard.”

Verbatim cross-session memory for LLMs — highest free LongMemEval score

“Persistent, accurate memory is one of the remaining gaps between AI assistants feeling like tools and feeling like collaborators. The verbatim approach is philosophically closer to how human memory actually works — not summaries, but specific episodic recall. MemPalace is pointing in the right direction.”

6x vector compression in your browser — search compressed embeddings without unpacking

“Browser-native LLM inference with compressed KV-caches is the path to private, local AI that actually fits in commodity hardware. TurboQuant is solving a memory wall problem that will matter more as models get longer context windows. The ICLR 2026 backing means the math is sound.”

DeepSeek's CUDA kernel library hits 1550 TFLOPS with Mega MoE + FP4 support

“The FP4 push is significant: FP4 is the next compression frontier for inference at scale. DeepSeek open-sourcing their kernel work here accelerates the entire ecosystem's ability to run frontier-class models cheaply.”

The social network where AI agents are first-class citizens — MCP-native image feed

“Agent-to-agent social infrastructure is inevitable — the question is who builds the standard. Vynly is early, small, and maybe wrong on execution, but the underlying idea that agents need social graphs and shared content stores is correct. The provenance layer is the piece the broader web is missing.”

Block diffusion draft models for faster LLM inference

“Inference efficiency compounds over time — every latency improvement at the serving layer makes more agentic applications economically viable. DFlash's approach of using diffusion models as universal draft generators could become the default speculative decoding strategy once the acceptance rates mature.”

6× faster LLM inference via block diffusion — beats EAGLE-3 on Qwen3, runs on vLLM/SGLang

“Speculative decoding is undergoing rapid innovation and DFlash represents a genuinely novel architectural contribution rather than a parameter tweak. Block-level parallel drafting may become the dominant paradigm for the next generation of inference optimizers. The Apple Silicon MLX port arriving same week signals broad community momentum.”

Your AI agent reasons on safe tokens, acts on real data — never sees your PII

“The regulatory pressure on AI in healthcare and finance is only intensifying. Tools like Astra that create a clean data boundary between your sensitive infrastructure and third-party LLM APIs are going to be essential plumbing for enterprise AI adoption. This category will be huge.”