Productivity / Research & writing
Research & Writing Copilot
Research you can trace, writing ready to publish.
Research & Writing Copilot turns sources, papers, notes, and drafts into technical articles, documentation, reports, and literature reviews whose key claims trace back to evidence. It shows uncertainty instead of hiding it, and never invents citations.
01 — The product
What is Research & Writing Copilot?
It’s for people who write technical and knowledge-heavy content that needs to hold up to scrutiny. It works out the task, whether deep research, a literature review, a fact-check, a refresh of an older piece, a comparison, or a full article.
Fluent prose isn’t the goal. Every important claim is labeled as verified, source-reported, inference, opinion, contradicted, stale, or unverified. Conflicting sources are compared by date, version, and scope rather than averaged into one number.
Research and presentation stay separate. One researched core can become a Medium story, a LinkedIn article or post, a developer blog, or a documentation page, and a final check confirms shorter versions don’t overstate the facts.
02 — Real workflow
Turning several sources into a grounded article
An engineer has notes from product docs, a vendor benchmark, and a forum thread, and wants a practitioner article comparing two caching approaches. The sources don’t fully agree.
01 · What you share
Topic: write-through vs write-behind caching for a read-heavy API Audience: backend engineers, practitioner level Source A: official docs for the cache product (version 7) Source B: vendor blog benchmark claiming 3x faster writes with write-behind Source C: forum thread reporting data loss on node failure Target: company blog, then a LinkedIn post
02 · How it works through it
- 1Defines the questions the article must answer and the claims that need verification before any drafting starts.
- 2Builds a claim ledger tying each claim to its source, source type, version, and status. The vendor benchmark stays a source-reported claim, not a verified fact.
- 3Drafts for the stated audience: problem, mechanism, example, trade-offs and failure modes, then a decision rule, with citations next to the claims they support.
- 4Runs a pre-publication review of sources, examples, and clarity, with findings marked blocking, important, or optional. Then it adapts the core article into a LinkedIn post without changing the facts.
03 · What you get
A publish-ready technical article with inline sources and visible caveats, plus a LinkedIn version built from the same researched core.
Illustrative workflow. It doesn’t depict a real session, system, or result.
03 — Areas of expertise
Where it helps
Fact-checks and refreshes
Marks claims in a draft or older piece as stale, conflicting, or unsupported, patches only those spans, and keeps the author’s voice. Untested code is labeled as such.
Literature reviews
Verifies titles, authors, DOIs, and publication status, extracts design and limitations into an evidence matrix, and separates peer review from preprints and vendor reports. No invented results.
Decision reports
Compares products or technologies on criteria that matter to your decision, normalized by version and tier. Recommends only when evidence supports it, and says what would change it.
Documentation and how-tos
Picks the right document type, from concept and tutorial to reference, migration guide, or runbook, and writes procedures with prerequisites, ordered steps, expected checks, and rollback.
04 — Working with Research & Writing Copilot
What to provide and what to expect
Share
- The topic, the question to answer, and the intended reader
- Your source notes, links, papers, files, or an existing draft
- The target format or publication, such as a company blog, Medium, or LinkedIn
- Any real experience or results you want included, since it won’t invent them
Expect
- Research briefings with each claim labeled by evidence status
- Technical articles, explainers, documentation pages, and research reports
- Comparison reports and literature reviews with evidence matrices
- Fact-check change logs, publication reviews, and platform versions of one article
No built-in connectors, and it doesn’t publish. Checking current sources, reading papers, and testing code depend on tools in your chat. Without them, it works from what you provide and marks unverifiable claims.