Good Morning! Here's what I have for you in today's newsletter:
xAI connects Grok Bot to bank, card, and investment accounts
Anthropic makes Claude Code wrap up cleanly at the 5-hour limit
ElevenLabs gives university students free access to its full platform
Pick the right Claude Code effort level for every kind of task
4 new AI tools worth trying today
AI AGENTS
xAI added a Finance integration to Grok Bot that links bank, card, and investment accounts through Plaid, so a Bot can answer questions and help manage spending from account data.
Once linked, a Bot can check balances, break down spending, find recurring subscriptions, and review investment holdings, all from a chat message.
Access is read-only through Plaid, so Grok Bot never sees bank logins and cannot move money, and any account can be unlinked at any time.
The integration lives in the Grok Bot marketplace under "Finance" and needs Grok Bot 0.49 or newer, and privacy mode still controls whether finance conversations are stored.
Grok Bot runs routines on a schedule, so now you can have a Bot check for new subscriptions or flag unusual card spending overnight. The Bots can read everything but not directly make any transactions, so you can use it without zero stress.
AI TOOLS
Anthropic updated Claude Code to finish open edits at the 5-hour usage limit, using a small fixed allowance from the weekly limit before it stops.
Pro subscribers get one wrap-up per week, while Max and Team Premium plans get one for each 5-hour session limit, as the change rolls out.
Claude finishes the task if only a few steps remain. Otherwise it completes the current piece and at most two more, then reports what is done and what is left.
The allowance also covers any subagents already running, and users who want to keep going past the wrap-up can switch to extra usage.
Sessions now end on finished edits with a note on where Claude stopped, so users can pick the work back up as soon as the limit resets. Max and Team Premium users get that safety net every session, while Pro users get it once a week.
AI MODELS
ElevenLabs launched ElevenLabs for Students, giving university students three months of ElevenCreative, ElevenAgents, and the ElevenAPI, plus a full year of ElevenReader Ultra, at no cost.
The plan is open to university students in the US, Canada, the EU27, the UK, and Australia, with more countries to follow.
ElevenCreative covers film, podcast, and social production, and ElevenAgents lets students build and deploy their own agents, such as research assistants and automation tools.
The ElevenAPI gives programmatic access to ElevenLabs' voice, sound, and music models for apps, hackathon projects, or a startup MVP, and ElevenReader Ultra turns reading lists into audio.
Three months of the full platform is enough time to build a project, from a narrated class presentation to a voice agent for a hackathon. ElevenReader Ultra runs for a year, so course packs and papers you'd rather listen to on the commute are covered for two semesters.
AI MODELS
Effort controls how much verification and edge-case testing Claude does, not how smart it is.
Step 1: Learn what effort does
Think of being asked to finish in 12 hours versus 1 hour. More time means more verification and independent judgment, not a smarter model.
Step 2: Build on low, verify on high
Have Claude interview you about your spec, build on low effort, review it, then verify on high.
Interview me about the details of this feature before you build anythingStep 3: Pick the level for the task
Low for quick, in-the-loop work like sketching. Medium for most feature work. High when verification matters. Max for fully autonomous, difficult problems.
/effort lowStep 4: Save high effort for edge cases
Higher effort cuts failures from missed edge cases, but it doesn't fix an approach that's wrong from the start. On an HTML sanitizer task, Fable 5.1 went from 1/5 at low effort to 5/5 at xhigh.
Build this and verify it against edge cases before calling it done
Step 5: Switch levels when needed
Change the level in the same conversation, no new chat needed.
/effort highP.S. You can access all the AI courses and trainings (including the full version of this one), prompts and workflows if you upgrade.
Midjourney released a new --tile for seamless patterns in v8.1 and v8.2 and a new inpainting model that changes only the selected pixels, replacing omni-reference, and opened an early preview of real-time models on its alpha site.
Notion added column permissions to databases, its most requested feature, so workspace owners can set exactly who can view and edit each property and keep sensitive fields like salaries or client notes private inside a shared database.
Lovable opened Chat mode for free, with a daily allowance on Free, Pro, and Business workspaces for planning an idea, turning feedback into a roadmap, or reviewing a live app before any code changes are made.
🎬 Runway: added Draft mode for Seedance 2.5, which generates faster at fewer credits while exploring, then enhances favorites to full quality.
📺 fal: US-hosted Seedance 2.0 now generates 1080p and 4K video, following last week's 1080p rollout for US-hosted Seedance 2.5.
⚡ MiniMax: added M3.1 Flash Preview to the Token Plan, built for high-volume, latency-sensitive workloads under an existing subscription.
💼 Slack: Rippling AI now works in Slack, answering HR and IT questions from live data and starting actions like payroll updates from a message.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial
✓ $500+ worth of AI tools