Edition W1 · Weekly
The complete week, consolidated.
JUNE 23–29, 2026. Seven daily editions in one place, with related coverage combined.
Top signals
10 signals
The week’s anchor
OpenAI previews GPT-5.6 Sol with restricted cyber-focused rollout
OpenAI previewed GPT-5.6 variants Sol, Terra and Luna to a small set of government-approved companies, with broader availability planned in the coming weeks and a July Cerebras launch mentioned by Andrew Feldman. Sol is positioned as the flagship cyber-capable model with stronger safeguards for sensitive cyber requests and reported performance competitive with Anthropic Mythos Preview on ExploitBench.
thehackernews.comRead the source →Gemini 3.5 Flash adds native computer use for browser, mobile, and desktop agents
Google DeepMind says computer use is now a built-in tool in Gemini 3.5 Flash, moving the capability from a standalone Gemini 2.5 computer-use model into the main Flash API and Gemini Enterprise Agent Platform. The model can see, reason, and act across browser, mobile, and desktop interfaces for software testing and knowledge-work automation.
Claude Code and Amazon Q bugs show the repo-trust problem for coding agents
Two disclosures showed AI coding tools turning repository-controlled state into executable behavior. A Claude Code sandbox escape used prompt injection and Git worktree path confusion to overwrite `~/.zshenv`, while Amazon Q Developer patched a high-severity MCP config flaw where a malicious `.amazonq/mcp.json` in a trusted workspace could launch local processes inheriting AWS credentials.
OpenAI says it built its first AI chip with Broadcom
OpenAI announced Jalapeño, its first in-house AI chip, saying it was designed with Broadcom for the LLM workloads behind ChatGPT, Codex, the API, and future agentic products. The company did not disclose specs, deployment timing, or whether the chip targets training, inference, or both.
Ornith-1.0 coding models land in Ollama with MIT-licensed open weights
Ollama posted run commands for Ornith after the Ornith-1.0 family was described as 9B Dense, 31B Dense, 35B MoE, and 397B MoE agentic coding models released under the MIT license. The quoted release claims strong coding results across Terminal-Bench and SWE-Bench variants, with local testing reported for the 35B MoE at FP8 on one DGX Spark.
OpenRouter publishes continuous GPQA and TAU-Bench results for open-weight models
OpenRouter says it continuously runs GPQA and TAU-Bench on most open-weight models and uses those results in AutoExacto, its default meta-benchmark for routing tool calls. Its GLM 5.2 page exposes 1M-token context, reasoning-effort settings, OpenAI-compatible API access, and rolling 30-day provider price, latency, throughput, success-rate, and eval charts.
vLLM shares a four-node DGX Spark recipe for Nemotron 3 Ultra 550B
vLLM highlighted NVIDIA’s guide for serving Nemotron 3 Ultra 550B-A55B-NVFP4 across four DGX Spark machines as one OpenAI-compatible endpoint. The setup uses tensor parallelism size 4, vLLM’s nightly aarch64 OpenAI container, RoCE networking, a 262,144-token max context, Mamba cache flags, and optional MTP speculative decoding.
Figma and Genspark push design tools closer to runnable code
Figma’s latest update adds code layers on the collaborative canvas, repo cloning, code-to-design flow extraction, animation and 3D transform support, AI shader fills, prompt-created plugins, and assistant skills that can use context from tools like Notion, Granola, Excel, and GitHub. Genspark separately launched Genspark Design for generating UI prototypes, videos, HTML animations, posters, and code.
Agent runtime tooling converges around harnesses, chat surfaces, and Claude Code orchestration
Vercel’s AI SDK Harness added adapters for LangChain Deep Agents and OpenCode, letting apps swap among Claude Code, Codex, Deep Agents, OpenCode, and Pi through one interface inside Vercel Sandbox. CopilotKit’s OpenTag repo offers a self-hosted alternative to Claude in Slack, while the Claude Protocol adds lifecycle hooks, task worktrees, logged decisions, and PR-based task isolation.
Amazon commits another $13B to AI and cloud infrastructure in India
Amazon says it will invest an additional $13 billion through 2030 to expand AWS data center capacity in Mumbai and Hyderabad, bringing its stated India commitments to $48 billion. The announcement follows major India infrastructure commitments from Microsoft and Google.
Tools & repos
10 pickscalesthio/OpenMontage
World's first open-source, agentic video production system. 12 pipelines, 52 tools, 500+ agent skills. Turn your AI coding assistant into a full video production studio.
google-labs-code/design.md
A format specification for describing a visual identity to coding agents. DESIGN.md gives agents a persistent, structured understanding of a design system.
DeusData/codebase-memory-mcp
A high-performance MCP server that turns codebases into persistent knowledge graphs is directly useful for agentic coding and code intelligence workflows.
opendatalab/MinerU
Converting complex documents into LLM-ready Markdown and JSON is a foundational capability for agentic document pipelines.
apple/container
A tool for creating and running Linux containers using lightweight virtual machines on a Mac. It is written in Swift, and optimized for Apple silicon.
mukul975/Anthropic-Cybersecurity-Skills
817 structured cybersecurity skills for AI agents mapped to MITRE ATT&CK, NIST CSF 2.0, MITRE ATLAS, D3FEND, NIST AI RMF, and MITRE F3, with support for Claude Code, GitHub Copilot, Codex CLI, Cursor, Gemini CLI, and other platforms.
Robbyant/lingbot-map
A feed-forward 3D foundation model for streaming scene reconstruction is relevant to robotics, spatial AI, and real-time perception work.
altic-dev/FluidVoice
A fully local macOS dictation app is relevant to developers interested in offline voice-to-text and privacy-preserving AI utilities.
commaai/openpilot
openpilot remains a major open robotics and driver-assistance codebase for engineers studying production autonomy systems.
ripienaar/free-for-dev
This long-running free-tier catalog remains a practical resource for developers and founders assembling low-cost infrastructure.
Blogs
6 readsvLLM-Omni details TTS serving optimizations for Qwen3-TTS, VoxCPM2, Higgs Audio V3, and Fish Speech S2 Pro
From x.com.
Google Cloud publishes ADK Skills codelab for deploying agents to Agent Engine
From x.com.
TPU architecture writeup explains systolic arrays, large on-chip memory, and pod-scale design
From x.com.
The Thermodynamic AI Chip · Thomas Ahle
From youtube.com.
HP Inc. launches Frontier strategic partnership with OpenAI
From openai.com.
Midjourney wants to delete 30% of all death...
From youtube.com.
Funding & acquisitions
13 movesGroq
Groq confirmed a new round while emphasizing a neocloud inference business spanning 13 data centers, five million developers, and trillions of tokens per week.
Sail Research
Sail Research is building an inference platform for long-running agents across heterogeneous hardware, custom kernels, and primitives for pausing, resuming, and forking whole VMs.
Patronus AI
Patronus AI raised funding to build digital worlds that replicate websites and internal systems for stress-testing agents after reinforcement learning.
General Intuition
General Intuition raised funding at a $2.3 billion valuation to train agentic models on gameplay clips with action labels from Medal.
Hang Ten Systems
Former Infosys CEO Vishal Sikka launched Hang Ten Systems to use AI-driven development and automation for building, modifying, and operating enterprise software.
Netris
Netris raised funding for switch-level automation that helps AI neoclouds configure, operate, and isolate GPU cluster networks faster.
Anduril
Anduril was highlighted as an AI-native defense platform around Lattice OS for sensor fusion, AI targeting, and autonomous coordination after a major Series H round.
MoEngage / Aampe
MoEngage acquired Aampe to add technology that creates a dedicated AI agent for each customer instead of relying on broad marketing segments and rule-based campaigns.
Adobe / Topaz Labs
Adobe is acquiring Topaz Labs and plans to integrate its AI image and video enhancement models into Firefly and Adobe creative apps while keeping Topaz products standalone.
BrowserAct
Web browser automation for AI agents.
Persona.js
WebMCP-native AI chat for frontends could help developers embed agentic interfaces directly into apps.
Figma Motion
Your Figma canvas now has a timeline.
Cloud World Model
A cloud simulation tool for AWS, GCP, and DigitalOcean could help engineers prototype infrastructure decisions without burning real spend.
Bengaluru radar
16 eventsReactNexus
An in-person React, React Native, and AI-for-frontend conference is a strong fit for frontend engineers and product teams.
Meesho x Elevation Capital: An evening with YC Alumni
A founder-focused fireside with Meesho and Elevation Capital offers useful perspective for YC-linked founders building at the frontier of AI and startups in India.
Bangalore Paper Club - Edition #1
A small-format research meetup covering LLM safety, edge inference, and LLM-based retrosynthesis is unusually high signal for ML engineers and researchers.
Developer Days Bengaluru
Workday's developer event offers hands-on labs for platform apps, agentic AI use cases, APIs, SDKs, and marketplace paths.
DevRev Leadership Circle Bengaluru
A senior-leader forum on operationalizing AI inside enterprise workflows is relevant for executives building AI into search, decisions, and execution.
AI Connect Season 2: The Next Chapter of Application Security
This session tackles how frontier models and autonomous agents are changing vulnerability discovery, exploit generation, and modern AppSec defenses.
Pipes, Streams, and Queries: Engineering Fast Data at Scale
ClickHouse and GlassFlow engineers will cover real-time pipelines, OpenTelemetry streams, Postgres migration, and low-latency analytics patterns.
Building Enterprise AI Agents with RAG using UiPath SDK and LlamaIndex
A practical session on coded agents, RAG, LlamaIndex, UiPath SDKs, and reusable skills is valuable for teams building enterprise automation with LLMs.
Inside High-Scale Systems: From AI-Driven Quality to Distributed Scheduling
The agenda pairs AI-powered quality engineering for large open data platforms with PhonePe-scale distributed scheduling for billions of callbacks.
Meetup: Memory, Context & Agents (SurrealDB x CCCL)
This deep dive on persistent memory, context layers, graph reasoning, and SurrealDB-backed agents addresses one of the hardest parts of production agent systems.
Gemma x Hugging Face Bengaluru Meetup
Google DeepMind and Hugging Face are running a hands-on builder session around the Gemma open model family, making it a strong event for practical open-model work.
Beyond GenAI: Crafting the Future of Customer-Facing Enterprise Applications
This mini hackathon with Amadeus and Contentstack focuses on moving beyond GenAI experiments into reliable, customer-facing enterprise AI applications.
Production-Ready AI: Observability, Scaling, and Best Practices AgentCore
This session focuses on monitoring, tracing, scaling, and debugging AI agents in production with OpenTelemetry-style observability.
Agentic Summit BLR
A dedicated showcase for agent products and infrastructure is highly relevant for engineers, founders, enterprises, and investors tracking the agent ecosystem.
Red Hat PyTorch x Hugging Face Bengaluru Meetup
A technical meetup covering SGLang, TRL reinforcement learning, PyTorch profiling, distributed training, and GPU-to-GPU communication.
DataHack Summit 2026
Analytics Vidhya's large AI conference focuses on human-AI collaboration, agentic systems, live builds, workshops, and industry AI applications.
Inbox edition
Know what matters before your day gets noisy.
Subscribe to Ekloge for one carefully curated AI briefing in your inbox—no endless feed, no filler.