Updates | AI Navigate

A · Theme of the day

Coding AI hits headwinds on three fronts

Security, regulation, and a Big Tech ban all landed on coding AI today.

Claude Code can be hijacked via a poisoned repo

Compared to before

Until last month, the pitch was 'hand it a repo and it runs.' That same auto-import step is now confirmed as an attack surface.

What changed

Claude Code runs hidden malware in GitHub repos without verification, handing attackers full control (THE DECODER). Auto-import is the attack surface.

Why it matters

Anyone running Claude Code in CI or locally should sandbox or add commit-signature checks now. Personal, low-volume use is lower risk.

Source: docs.anthropic.com　·　→ View on chaos map

Read in depth →

Meta bans Claude Code and Codex from its engineers

Claude Code

Compared to before

Last month, individual engineers at Big Tech could still quietly use external coding agents. Now company policy is getting ahead of it.

What changed

Meta blocked Claude Code and Codex to stop rivals training on Meta's data (THE DECODER). Big tech cutting off frontier coding agents is becoming a pattern.

Why it matters

PM and procurement teams selling into large enterprises: expect data-policy clauses in contracts to tighten. Any cloud-based agent faces the same exposure.

Source: docs.anthropic.com　·　→ View on chaos map

Read in depth →

GPT-5.6 delayed by U.S. government review request

GPT (OpenAI)

Compared to before

On 6/27, per-customer approval rules were already reported. Now the release date itself is in the government's hands — a new layer on top.

What changed

OpenAI reportedly delayed GPT-5.6 after a U.S. government review request (Reddit). Regulators are now influencing not just how but when frontier models ship.

Why it matters

Teams planning around GPT-5.x timelines should treat the schedule as uncertain. Competitors may face the same requests — this isn't OpenAI-specific.

Source: openai.com　·　→ View on chaos map

Read in depth →

B · Theme of the day

Salesforce puts a price tag on 'resolved'

Agentforce put a dollar value on 'resolved' — per-outcome pricing goes official.

Agentforce prices AI agents by outcomes, not conversations

Salesforce Einstein / Agentforce

Compared to before

Until now, Agentforce billed per conversation — you paid the same whether the agent solved anything or not, making ROI hard to calculate.

What changed

Salesforce formally defined 'resolved' for AI agents and published per-resolution pricing (Reddit). Billing moves from per-conversation to per-outcome.

Why it matters

Teams modeling AI agent ROI can now run cleaner numbers. Watch how 'resolved' is defined in the contract — that's where disputes will happen.

Source: salesforce.com　·　→ View on chaos map

Read in depth →

C · Theme of the day

Grok 4.5 enters the real-world lab

Grok 4.5 is now live inside SpaceX and Tesla as a dogfood beta.

Grok 4.5 goes live inside SpaceX and Tesla

Grok (xAI)

Compared to before

Six months ago Grok was behind on multimodal. Now xAI is dogfooding the next tier model in live production at sister companies before any external release.

What changed

Grok 4.5 is in internal beta at SpaceX and Tesla. Musk says it's 'approaching Claude Opus' (Innovatopia) — a preview of its entry into the frontier race.

Why it matters

For teams evaluating Claude Opus or GPT-5.x: a third frontier option is forming. No external access yet — put it on the watch list, not the shortlist.

Source: x.ai　·　→ View on chaos map

Read in depth →

D · Theme of the day

AI companies start showing their cards

Anthropic published real usage data; Suno launched an artist fund.

Anthropic publishes real data on how Claude is used

Claude (Anthropic)

Compared to before

Until now, AI usage patterns were each company's black box — you had to infer from anecdotes. This is Anthropic voluntarily releasing concrete numbers.

What changed

Detailed Claude usage report: personal use spikes on weekends, women spend more time per session, high-wage tasks often run at night (GIGAZINE).

Why it matters

Useful for enterprise sales teams and marketers building use-case pitches. If competitors follow suit, the benchmark data will get genuinely interesting.

Source: anthropic.com　·　→ View on chaos map

Read in depth →

Suno launches Spark, a fund for indie artists

Suno

Compared to before

Suno has been defending label lawsuits over training data. Rather than fighting creators too, it's now trying to bring them to its side.

What changed

Suno launched 'Spark,' an indie artist fund, using its $5.4B valuation to build creator-community support while label lawsuits continue (Innovatopia).

Why it matters

A signal Suno is playing the long game for designers and marketers using AI music tools. Lawsuit risk remains — watch how artists respond.

Source: suno.com　·　→ View on chaos map

Past updates

A daily archive of changes actually applied to the site.

Updates for 6/30

Coding AI hits headwinds on three fronts

Salesforce puts a price tag on 'resolved'

Grok 4.5 enters the real-world lab

AI companies start showing their cards

Past updates