Shield AI and Zipline lead the week’s largest U.S. startup funding rounds
Shield AI secured $2 billion in equity and $500 million in debt financing. Zipline International raised $1.5 billion at a $21.5 billion post-money valuation.
November 6, 2026
Friday
- DevinProduct updates
Devin introduces Review 2.0 with cheaper, faster PR reviews
A typical Devin Review now costs $0.25-$0.75 per PR. Median review latency has fallen from 139 seconds to 94 seconds since August.
Why it mattersTeams can configure automatic reviews to run on every push and set a per-PR spend limit.
- The VergeProduct updates
Perplexity launches Teammates, always-on AI agents for work
Perplexity’s Teammates can handle multiple workspace roles using existing projects, skills, and memory. Users can track progress in the Perplexity app and interact with Teammates in Slack.
Why it mattersBuilders who want to use Teammates need Perplexity’s $200 / month Max plan.
Alpine Linux releases fixes for OpenSSH password authentication bypass
Alpine 3.21.10, 3.22.8, 3.23.8 and 3.24.4 include the fixed package. Updating the package requires a separate sshd restart to apply the fix.
Why it mattersIf you cannot update yet, set PasswordAuthentication no in your sshd configuration and restart sshd to close the hole.
GitHub adds repository support for ACCESSIBILITY.md statements
GitHub now surfaces ACCESSIBILITY.md statements on repository overview pages. New repositories now get an accessibility issue label by default.
Why it mattersFor a public repository, you can add or propose an accessibility statement through Insights > Community Standards using GitHub’s starter document.
Palo Alto Networks in talks to acquire Clutch Security for $500 million
Palo Alto Networks is discussing buying Clutch Security, which protects AI agents’ access to company systems, for $500 million in cash plus milestone-based payouts, according to a source.
- NotionProduct updates
Notion sets its next direction around shared context for AI agents
Notion plans to serve as a context system for outside agents, with its own agents maintaining that context. Notion says its MCP volume has grown roughly 60x this year.
Cuban accuses Facebook of running deepfake scam ads as Meta documents point to fraud revenue
Mark Cuban says Meta runs scam ads using AI copies of his face and voice. Reuters reported that internal Meta documents projected roughly $16 billion in 2024 revenue from ads promoting scams and prohibited products. Meta disputed the documents' portrayal of its fraud response and called the revenue estimate overly broad.
GitHub introduces a pull request Overview page in public preview
The new Overview page is available as an opt-in public preview, with rollout continuing through the weekend. Depending on permissions, developers can mark drafts ready for review or close pull requests from the header.
Why it mattersSelect the ✨ Try the Overview page button in a pull request header to join; the same menu lets you return to the classic overview during the preview.
Dropbox to acquire Tigris Data for $350 million
Dropbox announced a $350 million acquisition of Tigris Data to expand its AI infrastructure. The transaction is expected to close in the fourth quarter of 2026.
Why it mattersBuilders using tools and applications built for Amazon S3 can evaluate the planned DBX Agent Cloud object storage offering without purchasing other Dropbox services.
Stripe opens analytics data to agents through MCP and APIs
All Stripe users can now connect compatible agents such as Claude and Cursor to their analytics data through Stripe’s MCP or the Queries API and Reports API. Users can inspect and edit the SQL behind each API analysis to verify calculations.
Why it mattersBuilders can try the Queries API free through January 31, 2027, then pay per credit without a monthly subscription, while the Reports API will remain free.
ERNIE-5.5-Flash-Base released with 280B total parameters
ERNIE-5.5-Flash-Base is a sparse MoE language model with 280B total parameters and 14B activated per token. The checkpoint has completed pre-training but has not undergone mid-training or post-training.
Why it mattersWhen tokenizing inputs, keep add_special_tokens=True because the model requires a BOS token at the start of every input.
