Need to know
- OpenAI lifted ChatGPT Work and Codex usage caps after GPT-5.6 launch doubled peak traffic.
- Microsoft 365 Copilot will adopt GPT-5.6 as its default model across Word, Excel, PowerPoint, and Chat.
- Anthropic extended Claude Fable 5 free access through July 19 as competitive pressure from GPT-5.6 intensifies.
- BeyondTrust launched Pathfinder NHI Governance to extend privileged access controls to AI and non-human identities.
- Okta warned of an active vishing campaign targeting Microsoft Entra passkey enrollment across six industry sectors.
New Releases
OpenAI temporarily removed the five-hour usage cap on ChatGPT Work and Codex and reset counters after GPT-5.6's launch generated roughly twice the previous peak traffic.
- Traffic surge context: The volume spike was large enough that OpenAI reset limits multiple times over the weekend before making the removal permanent for now.
- GPT-5.6 Sol impact: Sol's token costs will count less against allowances going forward, making the flagship tier more accessible to heavy users.
Microsoft announced GPT-5.6 will become the default model across Microsoft 365 Copilot, replacing the prior default in Word, Excel, PowerPoint, Chat, and Cowork.
- Access path: Microsoft will route M365 customers through the OpenAI API alongside its own natively served models, preserving the existing commercial relationship.
- Scope: The change covers the full Copilot surface in Microsoft's productivity suite, not just Chat, making it the broadest enterprise rollout of GPT-5.6 outside OpenAI's own products.
BeyondTrust released Pathfinder NHI Governance, extending its privileged access platform to cover every non-human and AI identity in an enterprise environment.
- What it governs: Service accounts, API keys, AI agents, and machine identities are brought under the same policy and audit framework as human privileged users.
- Market timing: The release lands as Saviynt and Keeper Security shipped competing agent-identity controls on the same day, signaling NHI governance is becoming a crowded sprint.
GitHub made its overhauled pull requests dashboard generally available, giving developers a unified inbox to triage review requests, CI failures, and merge-ready PRs.
- Why now: The dashboard was designed specifically to handle the review bottleneck AI coding tools created; the share of developers skipping manual review of AI commits jumped from 10% to 40% in a single month.
- Features shipped: Saved views, per-repo filtering, and reorderable sections let teams adapt the inbox to their own prioritization logic rather than GitHub's defaults.
Funding
Semiengineering's Q2 2026 roundup shows 80 startups raised over $6B, with AI data center chips still dominating but edge silicon returning on physical AI demand — a rotation that signals enterprise deployment moving closer to the endpoint.
IBM and Red Hat commercially launched Lightwell, backed by their $5B open source security commitment announced in May, offering 6,500+ remediated dependencies via Lightwell Network and a limited-access Clearinghouse Premier for patch embargoes — a direct play for the supply chain security budget that compliance tools like Chainguard are already competing for.
Case Studies
Snappy roughly doubled its book rate for qualified opportunities after shifting outbound motion to ZoomInfo buying signals rather than static list outreach.
- The mechanism: Reps prioritized accounts showing real-time intent signals instead of sequencing cold lists, compressing the time between signal and first contact.
- Supporting data: HBR research cited in the case study found firms contacting leads within one hour were nearly 7x more likely to qualify them than those waiting longer.
Trending on X
- GPT-5.6 vs Fable 5 model wars Developers on X are debating whether GPT-5.6 Sol has closed the gap on Claude Fable 5, with the dominant take being that Anthropic's rapid promotional extensions signal defensiveness rather than confidence.
- Anthropic Fable 5 free extension optics The community is parsing whether Anthropic extending Fable 5 access to July 19 is a smart retention move or a sign of monetization pressure, with several threads arguing OpenAI's usage cap removal forced Anthropic's hand.
- AI companies explaining products badly Ethan Mollick's thread arguing that most users have no idea what current models can do in agentic coding contexts sparked wide agreement that labs are losing the activation battle through poor communication, not poor capability.
- Google Gemini 3.5 Pro delay fallout Leaked reports that Google scrapped the base model and restarted pretraining, pushing launch to July 17, are generating speculation that Google is systematically waiting to see competitor releases before finalizing its own.
- Meta Muse Image consent controversy Meta pulling its Instagram-based AI image generator fewer than 72 hours after launch drew sharp reaction from security and policy circles, with the consensus being that allowing any public account to be used as a reference without consent was an obvious legal problem that should have been caught pre-launch.