The 2025 State of AI
Self-hostable personal AI assistant with its own email, voice, and hands-on apps — you configure it, it acts.
Vellum earns a recommendation when you actually want an agent to do things: it sends email from its own address, places real outbound calls with its own voice, posts to X and Slack, and builds working web apps from a conversation. What it asks in return is technical comfort — credits at $1 = 1 credit, CPU/RAM computer tiers, storage caps, and a $10/mo platform fee on Super ($100/mo) and Ultra ($200/mo). Compare it to ChatGPT or Claude if you want zero-config chat, or to a self-hosted stack like Open WebUI plus your own tooling if you want full control without paying for compute tiers. Vellum sits between: open source, self-hostable, but usable on day one from the free Base tier.
Verified 8d ago · liveness 80/100 · cite: rightaichoice.com/tools/the-2025-state-of-ai
- Power users who want email, calendar and social handled by an agent
- Developers who want to self-host and inspect an open-source assistant
- People who prefer voice and proactive heartbeat nudges over prompt-poking
- Operators running one assistant across real apps instead of a chat tab
- People who want a zero-config chat app with nothing to manage
- Teams needing a cloud IDE or sandboxed code execution as the core product
- Buyers unwilling to track credits, computer tiers and storage caps
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Vellum if you want an agent that needs no metering, no compute tier and no credit tracking — its price is configuration, and credits at $1 = 1 credit run on top of the $30–$200/month plans.
Every AI response, web search and image generation burns credits at $1 = 1 credit, on top of your monthly plan, so a busy assistant can outrun its tier.
Free Base gets you started with no credit card, Mighty at $30/month fits a single consistent user, and Super $100/month or Ultra $200/month plus the $10/mo platform fee fit operators running the assistant across email, social and app-building. Self-hosting on your own hardware carries no platform fee, which beats paying a computer-tier premium if you already run infrastructure.
In short
The 2025 State of AI — Self-hostable personal AI assistant with its own email, voice, and hands-on apps — you configure it, it acts. Best for Power users who want email, calendar and social handled by an agent, Developers who want to self-host and inspect an open-source assistant, People who prefer voice and proactive heartbeat nudges over prompt-poking. Free to start; paid plans from $30/mo.
What's new in The 2025 State of AI
Checked 9 days agoAcross the latest 4 updates: 1 feature update and 3 changelog entries.
Vellum v0.12.6 released
Vellum shipped build 0.12.6 on September 29, 2026. Patch release following v0.12.5; no feature notes published for this build.
Vellum v0.12.5 released
Vellum shipped build 0.12.5 on September 25, 2026, four days after v0.12.3. No feature notes published for this build.
Vellum v0.12.4 released
Vellum shipped build 0.12.4 on September 23, 2026. Patch release between the v0.12.3 feature drop and v0.12.5.
Vellum v0.12.3: in-chat document creation, TinyFish search, Telegram groups
Assistant can create documents from the chat or Library; TinyFish search and page fetch added as tool providers; Telegram group support; computer-use screenshots persist in chat.
What people actually say about The 2025 State of AI — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
42 mentions across 4 sources (Hacker News, Product Hunt, Bluesky, Lemmy) · researched Jul 5, 2026.
Average across the 4 sources that answered — each source counts once, not each post.
- +Proactive heartbeat checks reminders up to 10 times daily.
- +Self-writing skills autonomously from memory and past interactions.
- +Real-world actions: send emails, make calls, post on social media.
- +Build web apps from conversation (e.g., habit tracker).
- +Native Slack rendering supports tables and images.
- −Extremely limited direct user feedback available online.
- −Name collision with other reports causes confusion in search and discussion.
- −Setup complexity may overwhelm non-technical users.
- −Reliability and support quality completely unverified by community.
- −No dedicated community forums or active GitHub issue tracker visible.
- • Pay-as-you-go inference credits can accumulate beyond subscription
- • Self-hosting requires own infrastructure and maintenance overhead
Viability Score
How well maintained and how widely used is The 2025 State of AI? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Assistant with its own email address and custom subdomain (Super and Ultra)
- Outbound phone calls in the assistant's own voice
- Voice input: push-to-talk or always-listening, multilingual speech recognition
- Live camera sharing during voice calls
- macOS Companion Surface with a global voice shortcut
- Android app on Google Play
- iOS Control Center integration
- Build real, interactive web apps from a chat conversation
- In-chat document creation and document editor with research and drafting
- Inbox reading, triage, reply drafting and summarization
- Slack send, channel read, thread summary and reply
- X (Twitter) posting, thread replies and post-performance reading
- Linear ticket creation, assignment and state changes
- Proactive heartbeat checks up to 10x/day
- Inspectable memory graph and Personal Knowledge Base
About The 2025 State of AI
Vellum is an open-source personal AI assistant that runs in the Vellum Cloud or on your own hardware, VPS, or private cloud. Unlike a chat tab, it has its own identity: an assistant email address and custom subdomain on Super and Ultra plans, its own phone presence for outbound calls, and connected accounts it can post to. From the chat you can ask it to pull weather, calendar and news into a morning briefing, read and triage an inbox, draft and send replies, post to X or Slack, move a Linear ticket, order on DoorDash, or build a working web app rather than a mockup. Voice is a first-class input — push-to-talk or always-listening, with multilingual speech recognition and live camera sharing during calls — and the macOS Companion Surface gives you a global voice shortcut. Recent releases added in-chat document creation, TinyFish search and page fetch as tool providers, Telegram group support, and persistent computer-use screenshots in chat. Each assistant gets a small virtual computer with a CPU/RAM ceiling and persistent disk sized to your plan; activity is metered in credits at $1 = 1 credit for things like AI responses, web searches, and image generation. The Base tier is free with no credit card, Mighty is $30/month, Super $100/month and Ultra $200/month, with a $10/mo platform fee on Super and Ultra only. Self-hosting carries no platform fee. It suits technical operators, developers and tinkerers who want autonomy and inspectable memory over a polished but closed chat app.
Behind the Verdict
The interesting thing about Vellum is that the assistant is not you. It gets its own email address and subdomain (Super and Ultra), its own phone voice for outbound calls, and its own accounts on X and Slack. That separation is what makes proactive behaviour possible: heartbeat checks run up to 10x/day, subagent roles carry guardrails, and the memory graph is something you can open and inspect rather than trust blindly. Skills are the second differentiator — they are generated from memory and self-improve, and you can install new ones from a library, write your own, or just ask and watch it work out a method. Combined with custom Ollama URLs for local model routing, that is a genuinely extensible system, not a prompt wrapper. The cost of that agency is configuration. Credits are pay-as-you-go at $1 = 1 credit and cover AI responses, web searches, and image generation; plans above the free Base tier start at Mighty $30/month, then Super $100/month and Ultra $200/month, with the $10/mo platform fee (subdomain, static IP, priority support) applying only to Super and Ultra. Computer size is a ceiling, not a burst — Small is fixed on Base, while paid tiers open Medium, Large and XL. Storage grows online without restart but is still capped by tier. Where Vellum fits: power users automating inbox, calendar and social work; developers who want to self-host and read the source; anyone who prefers voice and heartbeat nudges to prompt-poking. Where it does not: teams that want a cloud IDE or sandboxed code execution as the core product, buyers who don't want to track credits and compute tiers, and people who want voice-first interaction with no setup at all. The docs are unusually blunt about the trade-off — your prompts do reach a cloud AI provider, and every sensitive action raises an Allow/Deny prompt with a risk badge unless you write a standing trust rule. That honesty is worth more than a feature list.
Researching The 2025 State of AI? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas The 2025 State of AI actually fits — and what changes day-one when you adopt it.
You set up the free Base assistant, connect email and Google Calendar, and ask for a morning briefing. Over the next week you wire heartbeat checks to triage the inbox and draft replies from the assistant's own address.
Outcome: Inbox triage and a daily briefing run without you opening a chat window, and you decide whether the credit burn justifies moving to Mighty at $30/month.
You clone the open-source assistant, point it at a private cloud host, and set a custom Ollama URL so requests route to a local model. You write a skill and add a standing trust rule for file access.
Outcome: The assistant runs with no platform fee and no computer-tier charges, and you control data, keys and the uptime SLA.
You connect Slack, Linear and X, then use subagent roles with guardrails to post deploy status, move tickets as work lands, and publish launch threads. Telegram group support handles the external channel.
Outcome: One assistant keeps four surfaces in sync, with permission prompts and the memory graph giving you a reviewable record of what it did.
Use Cases
- Automate a morning routine: weather, calendar and news pulled into a personalized briefing.
- Build a fully interactive web app such as a habit tracker through conversation.
- Connect an inbox (or use the assistant's own address) to triage, summarize and draft replies.
- Draft, post and reply on X, then read what people said about the post.
- Place outbound phone calls, such as rescheduling a dentist appointment, in the assistant's voice.
- Create, assign and move Linear tickets directly from chat.
- Post status updates into Slack channels and summarize long threads.
- Use the browser extension to read open tabs and act on the page you are viewing.
Models Under the Hood
as of 2026-10-10
Limitations
- Vellum is credit-metered: $1 = 1 credit and credits pay for what the assistant does, including AI responses, web searches and image generation.
- Paid tiers run Mighty $30/month, Super $100/month and Ultra $200/month, and the $10/mo platform fee (custom subdomain, static IP, priority support) applies to Super and Ultra on top of computer and storage add-ons, not to Mighty.
- Compute is capped by tier — Small 1 vCPU / 2 GiB at base, up to Medium, Large and XL — and persistent storage is capped per plan (10 GB / 30 GB / 60 GB).
- Each plan's listed computer size is a ceiling rather than a burst allowance.
- Self-hosting is open source and free of platform or computer-tier charges, but you supply the hardware and manage it yourself.
as of 2026-10-03
Verification history
We have re-verified The 2025 State of AI 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published The 2025 State of AI tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Base
$0/mo
Ideal for
Someone testing whether an agent that sends email and makes calls actually fits their workflow, without a credit card.
What this tier adds
Starting tier: free access with a fixed Small computer (1 vCPU / 2 GiB) and pay-as-you-go credits on usage.
Mighty
$30/mo
Ideal for
A single power user running the assistant consistently for inbox, calendar and social work every day.
What this tier adds
Adds a $30/month plan with 10 GB persistent storage and included Mighty usage, and avoids the $10/mo platform fee.
Super
$100/mo
Ideal for
An operator who needs the assistant to have its own email and subdomain, plus a Medium computer and more storage.
What this tier adds
Upgrades to a 2.5 vCPU / 5 GiB Medium computer, 30 GB storage, assistant email and subdomain, and adds the $10/mo platform fee.
Ultra
$200/mo
Ideal for
Heavy users running app builds, document creation and multi-channel automation who need the largest compute and storage ceiling.
What this tier adds
Upgrades to a 4 vCPU / 8 GiB Large computer and 60 GB storage on top of Super's email, subdomain and platform fee.
Custom Plan
Configure
Ideal for
Teams or individuals whose usage pattern does not map cleanly onto a fixed monthly tier.
What this tier adds
Configure a plan around how you work, with credits at $1 = 1 credit paid per use and the ability to cancel or change anytime.
Self-hosting
$0 platform fee
Ideal for
Developers who already run their own hardware, VPS or private cloud and want to own data, keys and the uptime SLA.
What this tier adds
Open-source deployment with no platform fee and no computer tier charges — you supply the infrastructure.
Where the pricing makes sense
The company stage and team size where The 2025 State of AI's pricing actually pencils out — and where peers do it cheaper.
Free Base gets you started with no credit card, Mighty at $30/month fits a single consistent user, and Super $100/month or Ultra $200/month plus the $10/mo platform fee fit operators running the assistant across email, social and app-building. Self-hosting on your own hardware carries no platform fee, which beats paying a computer-tier premium if you already run infrastructure.
Setup time & first value
How long it actually takes to get something useful out of The 2025 State of AI — broken out by persona, not the marketing-page minute.
Free sign-up takes about five minutes — the assistant wakes in your browser, you name it and start talking, no credit card required. Connecting email, Slack, Linear or DoorDash adds a one-time OAuth step each; self-hosting takes longer depending on your hardware or VPS, since you provision compute and configuration yourself.
Switching to or from The 2025 State of AI
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From ChatGPT or Claude: start on the free Base tier, keep your existing chat tool for drafting, and move inbox, calendar and social actions to Vellum once skills are configured.
- →From a self-built script stack: replace cron jobs and one-off API calls with heartbeat checks, skills and subagent roles so scheduled work runs inside one assistant.
- →From a hosted assistant with no export: rebuild context by talking to Vellum and letting it write facts into the Personal Knowledge Base and memory graph.
- →From another open-source assistant: deploy Vellum on the same VPS or private cloud — self-hosting carries no platform fee or computer-tier charges.
- ↗To a zero-config chat assistant: export your workspace, memories and config from the Vellum account, then paste the context you still need into the new tool.
- ↗To a full cloud IDE or agent sandbox: keep Vellum for personal ops while the IDE handles repository work, since Vellum's app building is conversational rather than a developer environment.
- ↗To fully self-managed open source: download for Mac or pull from GitHub and run the same assistant on your own hardware, keeping data, keys and uptime with you.
Integrations
Resources & Guides
- Documentationvellum.ai
Docs · The 2025 State of AI
Full product docs from vellum.ai
- Quickstartvellum.ai
What Is Vellum · The 2025 State of AI
Get up and running fast from vellum.ai
- Quickstartvellum.ai
Installation · The 2025 State of AI
Get up and running fast from vellum.ai
- Quickstartvellum.ai
Quick Start · The 2025 State of AI
Get up and running fast from vellum.ai
- Documentationvellum.ai
Get Started · The 2025 State of AI
Full product docs from vellum.ai
- Documentationvellum.ai
Faq · The 2025 State of AI
Full product docs from vellum.ai
- Documentationvellum.ai
Common Issues · The 2025 State of AI
Full product docs from vellum.ai
Tutorials & Learning
YouTube returned 6 videos for “The 2025 State of AI”, and we withheld 6: 6 did not mention The 2025 State of AI. We are showing none, because we could not prove any of them are about The 2025 State of AI.
Official links
Tools that pair well with The 2025 State of AI
Common stack mates teams adopt alongside The 2025 State of AI, with the specific reason each pairing earns its keep.
Lindy
Lindy is a Slack-native AI teammate that runs agentic workflows across 1,500+ integrations and answers your @mentions in-thread.
Poke (Interaction Co.)
Poke is an AI personal assistant that lives in your texts on Apple Messages, WhatsApp, Telegram and RCS
Superhuman
Superhuman is an AI email client, docs workspace, and Go assistant whose proactive agents run where your team already works.
Featured Head-to-Head Comparisons
The 2025 State Of Ai vs Locus Robotics
These tools are apples and oranges. Locus Robotics tackles physical warehouse productivity with proven AMRs and RaaS, while Vellum (its 2025 State of AI) is a personal digital assistant for automating online tasks. Choose Locus if you run a high-volume fulfillment center; choose Vellum if you want an AI that actually sends emails and makes calls for you.
The 2025 State Of Ai vs Truleo
Choose Truleo if you're a law enforcement agency needing to connect siloed data and automatically generate case leads, report writing, and jail call analysis. Choose The 2025 State of AI (Vellum) if you're an individual seeking a proactive, autonomous personal assistant that handles email, calls, and social media. They serve completely different domains — Truleo is a specialized investigation tool, Vellum is a broad personal productivity AI.
The 2025 State Of Ai vs Presto Voice
If you're a tech-savvy individual wanting a proactive AI assistant that manages emails, calls, and tasks autonomously, The 2025 State of AI is your pick. For QSR operators aiming to boost drive-thru revenue and efficiency with proven voice AI, Presto Voice is the clear winner—just expect enterprise pricing. Choose based on your domain: personal productivity vs. restaurant operations.
Alternatives to The 2025 State of AI
View allLindy
Lindy is a Slack-native AI teammate that runs agentic workflows across 1,500+ integrations and answers your @mentions in-thread.
Poke (Interaction Co.)
Poke is an AI personal assistant that lives in your texts on Apple Messages, WhatsApp, Telegram and RCS
Superhuman
Superhuman is an AI email client, docs workspace, and Go assistant whose proactive agents run where your team already works.
Frequently Asked Questions
Categories
Best-of guides
Used The 2025 State of AI? Help shape our editorial sentiment research.