Anthropic releases Claude Sonnet 4.6 with extended thinking mode and improved agentic performance. Available on claude.ai, API, and all major cloud providers.
Google launches Gemini 3.5 Flash at I/O 2026: fast, multimodal, $1.50/$9 per 1M tokens, powering the AI Mode experience in Search.
OpenAI's GPT-5.6 line spans Sol ($5/$30 per 1M tokens), Terra ($2.50/$15) and Luna ($1/$6), all with 1M context and effort-controlled reasoning — replacing the o-series as the default reasoning stack.
OpenAI releases GPT-5.5 at $5/$30 per 1M tokens. Native video understanding and a ~1M context window put it neck-and-neck with Claude Opus on coding.
Google I/O 2026 recap: Gemini powers Search AI Mode with 1M context, Project Astra persistent multi-modal agent, and major Vertex AI improvements for enterprise developers.
1.6T MoE model (49B active) under MIT license with a 1M context window. API access at $0.435/$0.87 per 1M — or fully open-source for self-hosting.
Amazon Bedrock expands its frontier model catalog with the latest from Anthropic and DeepSeek. Includes cross-region inference and updated Bedrock Guardrails.
Microsoft's Azure AI Foundry platform adds agent orchestration, built-in evaluation, content safety, and enterprise observability for production AI deployments.
Microsoft AI unveils MAI-Thinking-1 at Build — its first in-house frontier reasoning model family, in private preview on Azure AI Foundry. Reduces reliance on third-party model providers.
McKinsey's annual AI report: enterprise agent adoption tripled since 2024. 61% of large organizations run at least one production AI agent. ROI is finally measurable and positive.
Andrej Karpathy shares hard-won production lessons: context window engineering, eval design, cost budgeting, failure modes, and when NOT to reach for an LLM.
Simon Willison's running analysis of LLM deployment patterns: structured outputs, tool calling, prompt versioning, cost management, and the hallucination failure modes to watch.
Ethan Mollick argues AI literacy is as fundamental as spreadsheet skills were in 1990. Practical frameworks for integrating AI into daily professional work across every industry.
LangChain's engineering team explains how LangGraph handles persistent state, human-in-the-loop checkpoints, and parallel tool execution in real multi-step agent workflows.
Swyx argues that what you put in the context window — tools, memory, examples, constraints — determines output quality more than model choice or prompt wording.
Ben Tossell's weekly roundup: OpenAI's three-tier GPT-5.6 line, SpaceXAI's Grok 4.5 debut, full-duplex voice with GPT-Live-1, and the enterprise tools actually worth evaluating.
The Rundown's signal-vs-noise weekly digest: what's real, what's hype, and what every AI professional needs to know — model releases, tools, and enterprise adoption trends.
Accessible summaries of the most important AI research papers and product launches. Covers scaling results, agent benchmark progress, and policy developments from the past week.
What the EU AI Act actually requires from engineering and product teams: risk tiers, documentation obligations, prohibited use cases, and the audit checklist for high-risk systems.
Most-watched courses for each platform
freeCodeCamp.org
play_circleWatch
AWS Developers
play_circleWatch
Amazon Web Services
play_circleWatch
AWS Polska
play_circleWatchUpcoming AI conferences, summits & hackathons · Jul–Dec 2026
Empirical Methods in NLP — key venue for LLM and NLP research.
TAU's annual AI & tech week. Industry leaders, academic researchers, Google, NVIDIA. Runs alongside Cyber Week.
AI Security track — adversarial AI, LLM vulnerabilities, AI-assisted cyber defense.
The most prestigious ML research conference globally. Frontier AI, scaling, alignment.
Based in Israel? See the dedicated Hebrew board of local AI meetups, conferences and hackathons at אירועי AI בישראל.
