Swfte Journal / Ideas into practice

Perspective.
With a purpose.

Decisions behind the systems. Lessons from the work. Explore AI research, implementation guides and the ideas shaping what comes next.

334 articles
One connected library
Technology / Featured

Claude Fable 5.2: Release Date, $8/$40 Pricing and Breaking Changes

Fable 5.2 launched 9 November 2026 at $8/$40 per million tokens, down from $10/$50. What changed and who should switch.

·12 min de leitura
Read the story ↗

Explore the journal.

Find the perspective your next decision needs.

334 stories · Page 1 of 28

Technology
·12 min de leitura

Claude Fable 5.2: Release Date, $8/$40 Pricing and Breaking Changes

Fable 5.2 launched 9 November 2026 at $8/$40 per million tokens, down from $10/$50. What changed and who should switch.

Security
·11 min de leitura

AI Agent Identity and Least Privilege: Lessons From Anthropic

Anthropic reported four kinds of unintended agent action. What agent identity and least privilege should look like.

Technology
·10 min de leitura

Claude Haiku 5.5 Pricing: $0.10/$0.50 and a Cost-Per-Task Routing Test

Haiku 5.5 costs $0.10 in and $0.50 out per million tokens up to 100k. Cost per task, the 100k cliff, routing.

Guides
·11 min de leitura

EN 18286 Explained: The First EU AI Act Harmonised Standard

EN 18286:2026 is the first harmonised standard under the EU AI Act, for the Article 17 quality management system.

Technology
·11 min de leitura

GPT-6 Astra Edited a Model's Weights: What It Means for LoRA

MIT researchers say GPT-6 Astra fixed a small model by editing 0.016% of its weights. Evidence, critics, advice.

Guides
·7 min de leitura

How AI Agents Call Qwen 3.8: Tool Calling, Thinking Mode and MCP

Serve Qwen 3.8 with vLLM or Ollama, send tools, parse calls and fix raw XML. What is documented and what is not.

Strategy
·10 min de leitura

NVIDIA and Oracle Vera Rubin: What Multi-Gigawatt GPU Capacity Means

NVIDIA and Oracle plan multi-gigawatt Vera Rubin capacity. What is confirmed, what is a vendor claim, and what to do.

Guides
·12 min de leitura

Open-Weight Licences Compared: Beam, Mistral, Muse, Kimi, Qwen

We read the licence files. Beam is plain Apache 2.0. Mistral, Muse, Kimi and the larger Qwens add hosting terms.

Guides
·12 min de leitura

OpenAI API Alternatives After GPT-6.1 Terra and the Decisions API

Three kinds of OpenAI API alternative, a criteria table and a migration checklist after GPT-6.1 Terra.

Engineering
·12 min de leitura

Prompt Caching Pricing Compared: OpenAI 12-Hour, Anthropic, Qwen

OpenAI 12-hour cache writes cost 2.5x input. We work out break-even reuse for OpenAI, Anthropic and Alibaba.

Technology
·10 min de leitura

Qwen 4 Release Date: Not Out Yet, What We Know and Qwen 3.8 Today

Qwen 4 is teased but not released. What Alibaba has said, Qwen API prices today, and how to prepare.

Technology
·12 min de leitura

Reflection AI Beam: A 501B Open-Weight Model Under Apache 2.0

Beam: 501B parameters, 23B active, Apache 2.0. What is vendor-reported, what memory it needs and who should test it.

Browse every article (334)
The Swfte journal, by email

Fique à frente das tendências de IA

Receba insights semanais sobre IA empresarial na sua caixa de entrada.

Sem spam. Cancele a qualquer momento.

Ready to build with Swfte?

One platform for the agents, models and workflows your team ships. Free to start, no card required.