<nezhar/>

November 2026.2

November 9, 2026

This week: frontier models editing neural network weights, Claude Code leaking users' emails, Postman wiring itself into coding agents, Codex training opt-...

---

November 2026.1

November 2, 2026

This week a tampered Claude Code extension on Cursor's marketplace shipped an SSH backdoor to developer machines, Google says its new Gemini model tops the...

---

October 2026.4

October 26, 2026

This week AI shows both faces: frontier models help Wiz turn decade-old PHP flaws into remote code execution on WordPress, Drupal and Joomla sites, Kimi's...

---

October 2026.3

October 19, 2026

This week puts frontier labs under the microscope: security experts blame recent AI break-ins on skipped containment basics, Meta pitches Muse Wonder as a...

---

October 2026.2

October 12, 2026

This week covers three ends of the model market: Mistral previews Large 4, a trillion-parameter open-weight model; Cactus fits speech-to-text into 16.9 MB;...

---

October 2026.1

October 5, 2026

This week the focus is on the mid tier and on European open weights: Sonnet 5.5 and GPT-6.1 Sol both land at $2/$10 per million tokens while getting close to...

---

September 2026.4

September 28, 2026

This week the focus is on open alternatives to Jev and new model releases: Ollaya runs the open decision models locally, while Opus 5.5 and GPT-6 Sol and...

---

September 2026.3

September 21, 2026

This week is about trade-offs made explicit: Jev gives up text generation for millisecond decisions, Firefox trades local inference for a usable assistant,...

---

September 2026.2

September 14, 2026

This week is about AI rearranging business models: Tailwind joined Shopify after AI assistants dried up its docs traffic, DeepSeek made frontier agents...

---

September 2026.1

September 7, 2026

This week is about new releases, new benchmarks, and new prices: three frontier models in three days, benchmark numbers that change with the harness, and...

---

August 2026.5

August 31, 2026

Hey! Here's what caught my attention this week: what compliance costs small makers in the EU, and what open-source AI costs — Nvidia pays $13 billion for...

---

August 2026.4

August 24, 2026

AI keeps outgrowing its defaults this week: MCP replans the protocol around long-running agents, Qwen's new local model ships with a reasoning setting nobody...

---

August 2026.3

August 17, 2026

Artificial Analysis added six new models to its Intelligence Index this week: Grok 4.6 and DeepSeek V4 Pro at the frontier, Gemini 3.7 Flash in the workhorse...

---

August 2026.2

August 10, 2026

This week is about what you bring to the tool rather than how big the tool is: domain expertise turns out to be the real prompting skill, a 3B moderation...

---

August 2026.1

August 3, 2026

This week is about familiar ideas taking on new meaning in the agent era: an informal pelican benchmark gets tested for contamination, context windows become...

---

July 2026.4

July 27, 2026

This week is about boundaries—and what happens when they become unclear or fail completely. OpenAI’s agents escaped their sandbox and compromised Hugging...

---

July 2026.3

July 20, 2026

This week is about efficiency in different forms: making frontier AI cheaper to run, knowing when a side project has already served its purpose, and focusing...

---

July 2026.2

July 13, 2026

GPT-5.6 is finally public, Tencent released a much cheaper competitor, and a tool-calling bug showed how tightly models may be coupled to their preferred...

---

July 2026.1

July 6, 2026

Fable 5 is back, but its return shows how frontier model access now depends on policy, routing, safety classifiers, and credits. The rest of this week’s...

---

June 2026.5

June 29, 2026

This week connects three questions: where frontier AI is hosted, who gets access to it, and how we train the next generation of engineers in a world where...

---