Mistral is Europe's most successful AI startup, cutting through the market with a 'smaller, faster, cheaper' strategy and European data-sovereignty positioning. Mistral Large 3 is Europe's strongest commercial LLM, Small 4 is the 24B efficiency king, and Medium 3.5 is the open Modified-MIT model optimized for agentic coding. Its moat is not technical scale but the 'European compliance' card.
Qwen is the most-downloaded model family on HuggingFace, spanning sizes from 0.8B to 2.4T. In August 2026, Alibaba open-sourced a Max-tier flagship for the first time (Qwen3.8-2.4T-A95B) — but swapped the customary Apache 2.0 license for custom terms. Meanwhile the other new release, Qwen3.8-27B, runs native vision on laptop-class hardware and is the only one shipping under Apache 2.0. This post traces the family from 2023 through generation 3.8, explains how the open line and the commercial line split apart, and helps you pick the right model at each tier.
At Google I/O 2026, Antigravity CLI (agy) replaced Apache 2.0 Gemini CLI with a closed-source Go binary. Technical upgrades — multi-agent orchestration, native sandbox, millisecond startup — but free tier cut 98%, open-to-closed source, 28-day transition window. Community reaction was sharp.
Grok Build is xAI's Rust coding agent — 845K LOC, 8 parallel sub-agents, Arena Mode. May 2026 beta, July open-sourced (Apache 2.0) — but the direct trigger for open-sourcing was a privacy incident: it silently uploaded entire repos (including SSH keys, .env files) to Google Cloud Storage at a 27,800x traffic ratio. The exfiltration code remains in the binary, disabled only by a server-side flag.
In August 2026, Meta Superintelligence Labs released Muse Code beta. Closed-source static binary, Muse Spark 1.2 model, parallel persistent sub-agents with worktree isolation. The biggest controversy is pricing: Standard at $1.25/$4.25 per M tokens, or Contributor at $0.10/$0.20 — 20x cheaper, but your code enters Meta's training pipeline.
Xinference wraps vLLM, SGLang, llama.cpp, Transformers, and MLX under a single management layer, using a Web UI and OpenAI-compatible API to manage LLMs, embedding, rerank, speech, and image models — suited for self-hosted deployments that need multiple model types to coexist. But the management layer's parsing logic also creates a larger attack surface than pure serving engines (CVE-2026-61539 is a case study).
Agent Platform uses a Cloudflare-first architecture: local `npm run dev` runs Node-based simulations, while production maps to Workers + Workers Assets + D1 + KV + R2 + Vectorize + Queues + Workflows + Durable Objects + Workers AI. The Runtime interfaces stay the same (InMemory → Cloudflare implementations), so upper layers migrate without noticing. Deployment requires only `wrangler login` → create resources → fill in IDs → `wrangler secret put` → `wrangler deploy`. CI/CD watches the main branch and runs typecheck + build + dry-run + migration + deploy.
Evaluation is Agent Platform's quality immune system: instead of collecting statistics only after a run, it enforces checks throughout Pre-run, In-run, and Post-run execution. Seven eval categories cover Flow → Step → Skill → Artifact → Evidence → Policy → Regression. A Skill release must pass five gates—Trigger, Functional, Policy, Regression, and Human Review—and any failure blocks it. The Learning Loop moves from Run signals through Proposal, Human Review, Sandbox Eval, Quality Gate, and Publish, under one strict rule: agents propose, humans review, and eval gates decide whether a change can ship.
Flow Runtime is the heart of Agent Platform: a Flow becomes immutable when published, each Run is bound to a specific version and preset, Steps move through a DAG according to edge conditions, every boundary saves a checkpoint, and resume/retry-step preserves the complete trace history.
Observability is a first-class capability, not logging added after the fact: a structured trace connects FlowRun→StepRun→SkillInvocation→ProviderCall→ToolInvocation→GuardResult→EvidenceItem→ArtifactVersion. The Evidence Store traces every claim back to its source, excerpt, citation, confidence, and conflicts. Artifact versioning supports approve/reject/regenerate without deleting history. Context Snapshots allocate token budgets by category and record automatic compression when a block exceeds its budget. Procedural, episodic, and semantic memory can be written only through proposals reviewed by a human.
Agent Platform turns AI agents from a blank chat window into a structured workflow platform whose behavior can be defined, versioned, observed, verified, and improved. Its built-in Deep Research seed flow demonstrates the complete feedback loop.
The Policy Engine acts as the Agent Platform's constitution and enforcement layer: policies are versioned and bound to flows and presets; four guard layers enforce rules at step boundaries; budgets cap cost, tokens, runtime, iterations, and tool calls; external writes require human approval; loop detection trips circuit breakers; and escalation records provide an auditable trail. Rules are configuration-driven, so adding one means changing JSON rather than hard-coded logic.
The Provider Router is Agent Platform's model and tool gateway: it unifies 30+ providers, MCP tool discovery, step-local permission control, fallback chains with RRF fusion, and an OpenAI-compatible Proxy that existing SDKs can use without code changes. It is configuration-driven rather than hard-coded, with provider-health-aware routing.
A Skill is a versioned, installable, and auditable capability package. Its dual-file architecture separates metadata from instructions, explicit binding replaces model-driven routing, and every invocation is recorded. The Learning Loop turns run signals into proposals, sandbox evaluations, human review, and publication while enforcing the principle: agents propose, humans review, and evals serve as the gate.
Groundlane is an open-source TypeScript remote MCP server (v0.1.0) giving AI agents web_search, web_fetch, and web_extract through a single stable contract, with auth, provider routing, and resource limits kept at the operator boundary.
Hands-on parameter choices and response structures for web_search (ten adapters, RRF merge, dual-provider default), web_fetch (format/render strategies, finalUrl provenance), and web_extract (CSS selector determinism, no implicit LLM step), with verifiable error boundaries.
A four-dimension comparison (determinism, replaceability, identity boundary, operational cost) between Groundlane's controlled remote MCP contract and traditional local approaches (WebFetch, stealth_fetch, puppeteer, requests), with verifiable scenario recommendations.
Based on the in-site groundlane skill (mcp__groundlane__*) and usage-modes rules, this part describes reproducible steps for reference verification and digest data collection — without assuming unimplemented features or using deprecated stealth_fetch.
A reproducible checklist of the most common operational pitfalls: truncated results from fixed caps, selector errors tied to DOM stability, render-mode cost/determinism tradeoffs, version-change verification (v0.1.0 preview), and the non-negotiable security boundaries (URL policy, auth, concurrency, budget).
US-stock LLM agents have attracted nearly 100,000 GitHub stars, yet no Taiwan-stock project has even passed 10. I consolidated three side projects into a Taiwan-stock research agent where every conclusion must first survive a backtest; this article explains why.