OpenAI plans to launch its GPT-5.6 lineup next week, including standard, Mini, and Pro variants, with early Pro builds showing stronger long-horizon coding and a 1.5 million token context window. The move targets Anthropic’s top-tier models with aggressive token pricing, and a new bidirectional voice model, GPT-Bidi-1, is set to debut in ChatGPT alongside existing voice features.

• GPT-5.6 (standard, Mini, Pro) expected next Tuesday, with Pro's context window jumping from 1M to ~1.5M tokens and improved long-horizon coding via Codex
• Positioned to undercut Anthropic on price (already ~half Anthropic's token rates) while outperforming Claude Fable 5 on agentic coding tasks, timed to exploit Claude's regulatory troubles in the US
• New bidirectional voice model GPT-Bidi-1 can listen and speak simultaneously, handle interruptions, and adjust tone live, rolling out in ChatGPT with High/Medium/Instant tiers alongside a new draggable voice bubble UI

MIT CSAIL researchers built Retrieval Language Models that store full documents outside the model’s context window and let the AI query them via code, slicing, and parallel sub-instances. This approach handles inputs up to 10 million tokens, doubles benchmark performance, and matches or beats the cost of massive-context calls.

• MIT CSAIL's RLM approach stores full documents outside the context window and lets the AI query them via code (grep/slice) plus parallel sub-instances, instead of stuffing everything into context.
• It scales to inputs up to 10 million tokens, while standard LLMs like GPT-5 solve fewer than 10% of problems beyond 75,000 tokens.
• RLMs beat RAG methods by double-digit percentage gains at equal or lower cost per query.
• The code is open source and designed as a drop-in replacement for existing API calls, suggesting bigger context windows are a dead-end strategy compared to teaching models where to look.

This article shows that SKILL.md files aren’t static prompts but loader specifications defining what to load, when, and at what cost. It breaks down the three progressive-disclosure levels, explains how architecture—not instruction content—drives context consumption, and highlights common antipatterns that bloat or break skills.

• A 1,200-line monolithic SKILL.md ate 20% of the context window every turn; splitting it into a 180-line driver plus reference files and a script dropped that to 7%.
• SKILL.md files are loader specs with three tiers—frontmatter (~100 tokens, always loaded), body (loads on invocation, keep under 500 lines), and references/scripts (loaded or run only on demand).
• Adding YAML frontmatter to reference files accidentally promotes them to always-loaded skills, cluttering the skill list and causing out-of-context invocations—removing it fixed silent routing errors.
• Context consumption is driven by how content is architected across these layers, not by how much instructional detail is written.

Anthropic reduced Claude Code’s prompt cache TTL from one hour to five minutes, causing higher token write costs and faster quota depletion for long coding sessions. Developers report frequent cache misses—especially with large context windows—hitting usage limits and degrading performance. Anthropic says it will tweak default context windows but won’t offer a global TTL setting.

• Anthropic quietly cut Claude Code's cache TTL from one hour to five minutes in early March, causing far more expensive cache misses on long sessions.
• Cache writes cost 25% more per token than hits, so Pro users ($20/mo) report exhausting quotas after just two prompts in five hours.
• Anthropic says TTL is auto-selected client-side with no global override, but is testing a smaller 400K default context window (up to 1M optional) to curb costs.
• Beyond cost, users report degraded model behavior—looping, repetition, and "overthinking"—since a late-March update, separate from the caching issue.

This article breaks down the core concepts behind LLMs—from next-token prediction training to tokens, vectors and attention layers—to show how they generate text. It also covers context windows, parameters and why model scale affects performance.

• LLMs work purely through next-token prediction learned by hiding and guessing words billions of times during training, not through actual understanding of letters or math.
• Context windows now range wildly, from 200K tokens (~150K words) up to Llama 4 Scout's 10 million tokens, but bigger windows don't fix factual errors or logical gaps.
• Because models process text as tokens/vectors rather than raw letters, they inherently struggle with tasks like counting letters or doing arithmetic.