24 AUG 2026
Tingvar vs SynthBoard
Both argue that one assistant agreeing with you is the problem. SynthBoard argues it in writing, at length, and publishes real sessions. Tingvar argues it out loud and publishes the measurement instead.
Where they are ahead of us
SynthBoard publishes seventeen dated shared sessions, view counted, with the models named, read from www.synthboard.ai shared sessions index on 23 August 2026. That is more published evidence of its own behaviour than anything else in the crawled set, and we have published none, because zero real human sessions have taken place. If you want to read what one of these products actually produced before you pay for it, they can show you and we cannot.
The content surface is theirs as well. Around 331 pages, counted from www.synthboard.ai sitemap and a full crawl of it on 23 August 2026, against a register of 68 public routes here. A seventy two term glossary, read from www.synthboard.ai glossary index on 23 August 2026. Ours defines a term only where speech changes its meaning, so a reader looking for a vocabulary to read through has one there and not here. They own the anti-sycophancy search term outright, with a landing page and four posts behind it, and this page concedes the term rather than arguing about who should hold it. We will not match that content surface at launch. Their citation page for assistants was better than ours before we wrote ours.
They also ship a thing we have only planned. An agent protocol server available on every tier including the free one, read from www.synthboard.ai/mcp on 23 August 2026. Ours is on the roadmap and not in the product, so a team that wants to drive one of these from the tools it already runs can do that with them today and cannot do it with us at all.
The axis this comparison turns on
We concede the term. Arguing about who owns the word for a problem is not a use of anyone's afternoon, and they own it.
What is left is the mechanism. SynthBoard's product is text. Their single voice claim is one row in a feature comparison, available on the two most expensive tiers, with no page describing it, no sample, no demonstration and no published session in which it was used. Ours is a live room where one persona holds the floor, the room decides who, and your voice takes the floor from whoever has it.
And we publish numbers about it. Six metrics that would say whether the room does its job are defined on our evidence page with published targets and no measured values, because there is nothing to measure yet. That is a weaker claim today than an argument written well, and a much stronger one the moment it has a value in it. See what we have measured
Row by row
| Tingvar | SynthBoard | |
|---|---|---|
| Modality | Live audio room | Text. All seventeen published sessions render as written cards |
| Voice | The whole product. Four persona voice samples are published today | One feature row on a comparison table, on the two most expensive tiers, with no page, no sample and no published behaviour |
| Published sessions | None. Zero real human sessions have taken place | Seventeen, real, dated and view counted, with the models named |
| What separates the agents | What each optimises for, plus a published blind spot for each | Expert roles, described by domain |
| Published blind spots | Yes, one per persona, on /personas | Not published |
| Anti-sycophancy measurement | Six metrics defined, six targets published, no measured values | Not published. The argument is made in prose across a landing page and four posts |
| Published latency figure | Median 1471 ms, 95th percentile 1986 ms, from 122 readings in us-central1 on 19 August 2026 | Not published |
| Site size | 68 public routes in the register, and this site names the ones not written yet | Around 331 pages |
| Glossary | Six terms, only where speech changes the meaning | Seventy two terms |
| Machine readable page copies | Sixteen, and each one resolves | Published across the site, and forty of them return 404 |
| Pricing without JavaScript | Server rendered text | Their pricing page returns 200 and ships a loading placeholder to a client with scripting off |
| Tier badges | One tier carries a badge | Three of four tiers carry a superlative badge, which distinguishes none of them |
| Agent protocol available to your own tools | Planned, not shipped | Shipped, and available on every tier including the free one |
When to use each
Reach for SynthBoard
- You want to read a real session before you buy anything.
- You want a written record with a recommendation at the end of it.
- You want to drive it from the tools you already run.
- You would rather read than listen.
Reach for Tingvar
- You want to hear the disagreement rather than read a summary of it.
- You want to interrupt the moment something is wrong.
- You want the mechanism published as numbers, including the empty ones.
Questions about this comparison
What is Tingvar bad at?
Tingvar is bad at anything you want agreement or output from: it will not draft your deck, write your code, or summarise your meeting. It is a poor fit for casual brainstorming, for a decision you have already made, and for anyone who wants their instinct confirmed. It has no published customer outcomes, because zero real human sessions have taken place. The evidence page carries what we have measured and what we have not, including the rows that are empty.
How we compare
Every claim about another product on this site is a fact that product publishes about itself, read from its own pages on the date below. Where a product publishes nothing on a point, the cell reads not published, because an absent published fact is not an absent capability. Our own weaknesses are in the same table, in the same column style.
Compared on 23 August 2026
Sources
| Claim | Read on their own pages at | Read on |
|---|---|---|
| Text, with all seventeen published sessions rendering as written cards | www.synthboard.ai shared sessions index and the session pages themselves | 23 August 2026 |
| Voice as one feature comparison row, gated to the two most expensive tiers, with no marketing page, screenshot, demonstration or published session | The feature comparison data in the shipped www.synthboard.ai bundle | 23 August 2026 |
| Around 331 pages | www.synthboard.ai sitemap and a full crawl of it | 23 August 2026 |
| A seventy two term glossary | www.synthboard.ai glossary index | 23 August 2026 |
| Forty machine readable page copies returning 404 | Direct requests to each advertised twin on www.synthboard.ai | 23 August 2026 |
| The pricing page returning 200 and shipping a loading placeholder with scripting off | www.synthboard.ai/pricing, fetched with scripting disabled | 23 August 2026 |
| Three of four tiers carrying a superlative badge | The tier definitions in the shipped www.synthboard.ai bundle | 23 August 2026 |
| An agent protocol server available on every tier including the free one | www.synthboard.ai/mcp | 23 August 2026 |
Right of reply
If you work at one of these companies and something here is wrong, email corrections@tingvar.com. We correct it within 1 business day and log the correction publicly.
Correction log
No correction has been requested or issued. This log is dated and it stays here whether or not it has entries.
Listen to four voices first
The voice samples on the persona page are real. The words in them are scripted, and the page says so. Hear the four voices