Research Period: 17 August to 24 August 2026
Please check the infographic below.
Please review the slide version of the newsletter here.
1. Your GPT-5.6 Sol API bill just dropped, automatically
What changed
OpenAI cut the API and ChatGPT credit pricing for GPT-5.6 Sol by 20–33%, effective 21 August, with the promotion running through 21 November 2026. Input tokens drop from $5 to $4 per million (-20%); output falls from $30 to $20 per million (-33%); cached input also reduced. The lower prices apply to API access, ChatGPT Work credits, and Codex.
Why it matters
If you’re already using Sol via the API or ChatGPT credits, this saving is automatic: nothing to switch, nothing to set up. For teams running high-volume tasks, a 33% drop in output costs adds up quickly. Sol now costs less than Claude Opus 5 on both input and output, which makes it worth revisiting your model choice if you’ve been comparing options. Bear in mind this is a promotional rate through November, so it’s worth locking in workflows now and building usage data before standard pricing resumes.
How to use it
- Review your current OpenAI API or ChatGPT Work spend and update your cost forecasts for Q3 and Q4
- If you’ve been hesitating to migrate high-volume tasks to Sol because of cost, run a direct comparison now
- Use the promotional window to pilot workloads you weren’t running before, at lower cost
- Codex users: check your project budgets, as the same reduction applies to your coding tasks
- Compare Sol’s current pricing against other frontier models to inform your next renewal or contract decision
2. Perplexity now remembers everything, and gets better every time you use it
What changed
Perplexity shipped two Computer agent updates. First, forward or CC co******@********ty.com on any email and the agent runs the task, replying in the same thread. Second, Brain, a self-improving memory system, compiles a structured knowledge graph from your past work, so results improve on repeated queries: Perplexity’s own testing reports 25% better answer accuracy, 16% better recall and 13% lower cost on repeated tasks. The tiers differ. The email interface comes with Computer, which Pro subscribers have had since March; Brain is a research preview on Perplexity Max at $200 a month. Projects extends Brain across a team via Google Workspace and Slack.
Why it matters
Brain acts as a research assistant that reads your past work and gets better at it, so regular competitor, supplier or market research improves without extra effort, and the email route removes app-switching entirely. The caveat is price: email is within reach on Pro, but Brain sits behind the $200 Max tier, which is a decision rather than a switch to flick. For teams on Max, Projects means a shared research memory that compounds across everyone’s use.
How to use it
- Set up a recurring research task (competitors, industry news, pricing) and let Brain build context over the next few weeks
- Forward supplier research requests directly from your inbox to co******@********ty.com
- Create a Project for your team to share a common research context across Slack and Google Workspace
- On Max, run the same searches you’ve always done; Brain improves in the background with no setup required
- Track how answer quality changes over four to six weeks of consistent use, then share findings in the Guild
Sources: Perplexity Blog | AI Weekly
3. AI coding agents now live inside Slack: trigger, review, and merge without leaving the chat
What changed
Salesforce launched Slack Code, embedding AI coding agents directly into shared team channels. The agents supported at launch are Claude Code (Anthropic), Devin (Cognition), GitHub Copilot, and Vercel’s agent. The whole team can see the agent’s working conversation in real time, view live code diffs and HTML previews, leave feedback the agent incorporates, and approve the finished output; all without leaving Slack. Available on any Slack plan; you supply your own agent subscriptions separately.
Why it matters
For teams where the business owner needs visibility into what AI is building, this changes the dynamic. Instead of one developer running an agent in isolation and reporting back, the whole team sees the work as it happens. You can redirect the agent mid-task, catch problems early, and approve the result without a hand-off. It’s particularly useful for smaller businesses where non-technical staff and developers share the same Slack workspace daily.
How to use it
- Connect your existing Claude Code or GitHub Copilot subscription and try a low-risk first task from Slack (a documentation update or a bug description)
- Give non-technical team members visibility into AI coding work without granting them repository access
- Use the shared channel as a live review board: comment, redirect, and approve in one place
- Run an experimental project entirely from Slack to test whether it reduces your team’s context-switching
- Use the HTML preview feature to check front-end changes before they reach a staging environment
Sources: Salesforce/Slack | VentureBeat
Additional Noteworthy
ChatGPT for Teens: OpenAI rolled out a dedicated teen mode
ChatGPT Ads expand to 31 European markets: OpenAI extended its ad rollout to 27 EU member states plus Norway, Iceland, Liechtenstein, and Switzerland. UK members were covered in last week’s update; EU-based members can now see ads in ChatGPT Free-tier searches. | OpenAI
Cursor Origin, a GitHub alternative built for AI agents: Cursor launched a code hosting platform with repos, pull requests, and bidirectional GitHub sync, timed the same day as a significant GitHub outage. Privacy caution: it shipped without explicit data terms. Do not use with client code until those are published. | Cursor Changelog | VentureBeat
ChatGPT for Teens: OpenAI rolled out a dedicated teen mode
Keep exploring
More from the Guild.
AI Moments lands weekly. The Guild’s sessions and articles sit alongside it and go deeper on the things that matter.