…
Vol. 1·No. 1·Established 2026·kvire.com

Your senior
peer reviewer.

Paste a draft. Get peer-review feedback in seconds — anchored to verbatim quotes, calibrated to the evidence you actually wrote down, with the literature cross-checked against your claims. The prompts are engineered to journal-reviewer standards; the model is forbidden from inventing what isn't on the page.

Start writing What's inside

Free while in early access. No card, no install. Bring a draft of ≥ 800 characters; bring a PDF up to 25 MB.

I.

What Kvire actually does

Three capabilities, each shown below as it actually appears in the workspace. No mock-ups; the output format on this page is the output format you receive.

Exhibit A

Peer-review feedback

The structured response a senior reviewer would write to their editor. Five required sections, anchored quotes, no padding.

Central claim. Global temperatures have risen 1.2 °C since pre-industrial times, driven primarily by anthropogenic greenhouse-gas emissions, and existing policy commitments are insufficient to limit warming to the 1.5 °C target.

What's working. Precision on the scientific record: “risen by 1.2 degrees Celsius since pre-industrial times” grounds the opening in a verifiable figure.

What needs work. Causal language stronger than the cited evidence licenses; the leap from insufficient commitments to policy failure skips two intermediate steps.

Top three priorities. (1) Cite the IPCC AR6 finding directly. (2) Split paragraph 3 — it argues two things. (3) Replace “will fail” with the conditional the data supports.

Exhibit B

A line-level check

28 named fallacies in the canonical taxonomy. Each finding ships with a verbatim quote, a severity, and a one-line repair. Output contract is strict.

“Critics of universal basic income want people to starve rather than receive any government support.”
Fallacy
Strawman (canonical)
Why
Attributes an extreme position critics do not hold; the actual UBI debate concerns labor supply and fiscal sustainability, not mass starvation.
Severity
serious
Repair
Engage the steelmanned UBI critique — concerns about work-incentive effects — and respond to that.
Exhibit C

Evidence from the literature

250 million papers indexed. Pull supporting, contradicting, or related work for any claim in the draft. Every retrieved paper gets a one-sentence verdict.

Claim: “Remote work increases knowledge-worker productivity.”

  1. Does Working from Home Work? Evidence from a Chinese ExperimentBloom, Liang, Roberts, Ying · QJE · 2015 · 3,478 citationsSUPPORTS13% performance gain in a randomized trial of 249 Ctrip employees over 9 months.
  2. Work-from-anywhere: The productivity effects of geographic flexibilityChoudhury, Foroughi, Larson · Strat. Mgmt. J. · 2021 · 811 citationsWEAKLY SUPPORTSUSPTO patent examiners — 4.4% output increase, but only for the work-from-anywhere subset, not WFH.
  3. Working remotely? Selection, treatment, and the market provision of remote workEmanuel, Harrington · AER · 2023 · 204 citationsCONTRADICTSCall-center study: remote workers were 12% less productive than office workers in the same role.
II.

How it’s different from ChatGPT

Generic chat models will produce something plausible for any prompt. Kvire is engineered to produce something specific — or nothing at all.

28
named fallacies in the canonical taxonomy

From Affirming the consequent to No true Scotsman. The model can only report what fits the taxonomy; coining a new name is forbidden.

250M
papers in the corpus, anchored to every claim

The Evidence tab pulls live from OpenAlex, ranks candidates against the claim's actual wording, and writes a one-sentence verdict — supports, contradicts, or irrelevant.

0
inventions allowed — refusal sentinels mandated

When a check finds nothing, the model is required to say so verbatim. “No clear logical fallacies.” No padding, no inventing to fill the page.

III.

Engineered prompts, not vibes

The product is the prompts. Not the editor, not the sidebar. Below is the actual logic behind one of the most-used checks — published so you can audit it.

The taxonomy

The model is constrained to this list. Findings that don't match a canonical name are not reported.

Formal
  • Affirming the consequent
  • Denying the antecedent
  • Undistributed middle
Relevance
  • Ad hominem
  • Tu quoque
  • Genetic fallacy
  • Appeal to authority
  • Appeal to popularity
  • Appeal to emotion
  • Red herring
Induction
  • Hasty generalization
  • Cherry-picking
  • Anecdotal evidence
  • Survivorship bias
Causal
  • Post hoc ergo propter hoc
  • Cum hoc ergo propter hoc
  • Reverse causation
  • Single-cause fallacy
Structure
  • Strawman
  • False dichotomy
  • False equivalence
  • Slippery slope
  • Begging the question
  • Loaded question
  • Equivocation
  • No true Scotsman
  • Moving the goalposts

The output contract

Every finding ships in exactly this shape. The model is forbidden from emitting anything between findings — no preamble, no progress narration.

> "[verbatim quote from the draft,
   5–40 words, contiguous]"
**Fallacy:** [Canonical name]
**Why:**     [One sentence — name
              the broken inference.]
**Severity:** [minor | serious | critical]
**Repair:**  [One sentence — minimal
              fix or evidence needed.]
IV.

What you get

Six surfaces around one writing canvas. No mode switching, no plugin store, no setup.

I.
Actions

Eleven peer-review prompts: general feedback, fallacies, weak arguments, counter-arguments, contribution statement, limitations audit, methods clarity, and more.

II.
Checks

Eight line-level scans for passive voice, clichés, filler, excessive hedging, overconfidence, repetition, undefined jargon, and missing citation hooks.

III.
Evidence

Pick a claim, choose support / contradict / related — get back ranked papers from a 250M-document corpus, each with a one-sentence verdict.

IV.
History

Auto-snapshots while you write. Roll back to any point. No git, no commit messages, no friction.

V.
Chat

Multi-turn conversation about the draft. The model holds the full text in cached context and stays scoped to it.

VI.
Download

Export as PDF, Markdown, Word, or HTML. Your draft remains yours.

Open the editor. Paste a draft. Run a check.

Start writing

It takes about two minutes to know whether Kvire is what you've been looking for.

Kvire — your senior peer reviewer