June 25, 2026 (Thu)

[01]

Top Stories

1. OpenAI + Broadcom Reveal Jalapeño — First Custom AI Inference Chip

OpenAI and Broadcom unveiled "Jalapeño," OpenAI's first custom ASIC, on June 24, 2026 (OpenAI)
Design-to-tape-out in 9 months, with AI assistance — claimed fastest ASIC development cycle ever in high-performance semiconductors
Engineering samples already running GPT-5.3-Codex-Spark at production target frequency and power
"Performance per watt substantially better than current state-of-the-art"; initial deployment target: end of 2026
Accompanies a strategic collaboration targeting 10 GW of OpenAI-designed accelerators deployed with Broadcom
Why it matters: OpenAI is no longer solely dependent on NVIDIA for inference compute. Designing custom silicon signals a vertical integration play spanning model → chip → kernel → serving — the same strategy that gave Google (TPUs) and Amazon (Trainium) structural cost and performance advantages. This is the clearest sign yet that frontier AI labs are becoming full-stack infrastructure companies.
→ OpenAI (source)

2. Anthropic Launches Claude Tag for Slack — Shared AI Teammate for Teams

Claude Tag ships June 23, 2026 in beta for Enterprise and Team plan customers (Anthropic)
Places Claude as a shared, persistent AI teammate for entire Slack workspace channels — anyone types @Claude to delegate; Claude posts asynchronously, builds team memory, and can proactively surface information in "ambient mode"
Runs on Claude Opus 4.8; the prior Anthropic Slack app retires August 3, 2026 (30-day migration window for admins)
Karpathy (Anthropic pretraining team) noted June 24 on X: "I work from Slack now" — consistent with internal adoption
Why it matters: Claude Tag shifts the deployment model from per-user productivity to group-layer shared agent — the first Anthropic product designed around team shared context. This is the Slack-native realization of the Managed Agents vision, bringing ambient agentic AI to the communication layer where most work already happens.
→ Anthropic, Agents (LLM Agents) (source)

[02]

Paper Picks

No Tier-1 org papers above threshold from June 24-25. HuggingFace Daily Papers search returned non-Tier-1 orgs only (403 direct fetch, WebSearch fallback below interest threshold).

[03]

Watch

xAI Grok Build /goal mode (June 22) — long-running autonomous plan→execute→verify loop for Grok Build. Two-model pipeline (Composer 2.5 + Grok Build 0.1). SuperGrok/X Premium+ subscribers. Brings agentic coding task autonomy to a consumer subscription tier — watch for benchmark comparisons vs. Claude Code and OpenAI Codex. → Grok Build (source)
Google DeepMind + A24 deal (June 22, $75M) — Google's first equity stake in a film studio. DeepMind researchers embedded in A24 productions to build AI filmmaking tools (storyboarding first). SAG-AFTRA 2026 Agreement (ratified June 4) provides the protective framework. Watch for tool releases and DeepMind creative-AI expansion into Gemini Omni workflows. → Google DeepMind (source)
Gemini 3.5 Pro GA delayed to July 2026 — Google officially slipped the launch from June to July. Still in limited enterprise preview on Vertex AI. 2M token context + Deep Think mode still pending. Market prediction odds for "released by June 30" have collapsed. → Gemini 3.5 Pro

[04]

New in Wiki

No new pages created today (only updates to existing pages).

[05]

Updates

OpenAI: Added Jalapeño chip (June 24) to Recent Activity and Strategic Position; updated source list; updated → 2026-06-25
Anthropic: Added Claude Tag for Slack (June 23) to Recent Activity; updated source list; updated → 2026-06-25
xAI: Added Grok Build /goal mode (June 22) to Recent Activity; updated source list; updated → 2026-06-25
Grok Build: Added /goal mode section (spec, pipeline, significance); release date updated; updated → 2026-06-25
Google DeepMind: Added A24 deal (June 22) to Recent Activity; updated source list; updated → 2026-06-25
Gemini 3.5 Pro: Updated GA status to "delayed to July 2026"; updated → 2026-06-25