When it comes to finding people, the biggest hassle is never that you can’t find anyone—it’s that once you do, you don’t dare use it. You enter a name and get a dozen or more results, with plenty of people sharing the same name; you click through to check a résumé, and it looks quite polished, but there’s no telling where this “So-and-So University” actually is or whether this stint in a role belongs to the same person. In the end, you’ve usually saved a pile of links and not dared put a single one into formal materials.
MuchSci’s newly added Talent Profile takes a different approach: set the threshold first, then let the material in. This article won’t discuss how useful it is; it will just unpack one more interesting question—why it dares to treat what it has organized as usable records.
01 Before Entry, It Must Pass Three Gates
First, look at the most central design of this feature: the admission threshold.
▲ Figure 1 | It doesn’t accept everything it captures—the most important design of this feature lies entirely in the “entry threshold.”
The first gate is source scope. It only includes records related to national-level projects. This cut is crucial: once the scope narrows, the amount that needs verification drops, and the credibility of the sources rises across the board. MuchSci writes this into the header of the record, so users can see at a glance the scope within which what they’ve received was organized.
The second gate is three-way matching. The criteria are written quite strictly: a record is included if and only if the name, institution, and career history can be matched item by item and exactly against public materials. Note the phrase “if and only if”—matching the name alone doesn’t count; the institution and career history must match as well.
The third gate is isolating doubtful records. If the same project has inconsistent names across different materials, or the information doesn’t match, the record is marked “Pending Verification” and is not counted as an expert.
Taken together, these three gates answer the same question: why should this record be trusted? And the answer it gives is not “because I captured it,” but “because it can withstand item-by-item verification.”
02 A Record Is Assembled from Four Types of Fields
Records that pass the threshold are broken down into four types of fields.
▲ Figure 2 | From “Who is this person?” to “What have they done?”, each of the four categories answers one question
Basic Information answers “Who is this?”: a one-sentence bio, gender, date of birth, place of origin, current position, and research areas. Education answers “Where did they study?”, using a table that lists school, major, and result in chronological order. Employment history answers “Where have they worked and what have they done?”, also in a table, with columns for time, institution, and position. Main research achievements answer “What have they produced?”, divided into academic contributions and projects led, with projects involving major national engineering efforts specially marked.
The latter three categories all use tables rather than paragraphs—this is not a formatting preference. Each row in a table is an independent unit for verification—if a row looks suspicious, you can check that row on its own without rereading the entire paragraph. The real value of structuring is that it breaks the vague overall judgment of “whether to trust this person” into a set of small questions that can be checked one by one.
03 What’s the Difference After Structuring?
Put the two approaches side by side, and the difference becomes much clearer.
▲ Figure 3 | The difference is not in the number of words, but in whether it can be verified item by item and reused repeatedly
A single-paragraph introduction reads more smoothly, but it has three unavoidable weaknesses: verification can only be done as a whole—once you doubt a certain year, you have to read through everything again; sources are easily lost, and after a while you can no longer say where a piece of material came from; and it has to be rewritten for a different occasion—materials for a conference invitation and materials for a grant application often have to be written separately.
Structured records address each of these three points: items are listed individually, so a suspicious row can be checked separately; each record carries a data source identifier, so the source is not lost; and the same record can be reused in different scenarios without being rebuilt.
One point needs to be made especially clear here: structuring does not equal evaluation. A record organizes public materials; it records “where this person studied, where they have worked, and what they have led,” not “how impressive this person is.” The former is verifiable fact; the latter is a judgment that requires a person to make.
04 Feature Interface: What a Record Looks Like
The interface is under the “Talent Profile” tab of the agent assistant.
▲ Figure 4 | Basic information and education: a short bio plus a timeline, with the record identifier and data source written at the very top
The top line is this mechanism’s public statement: the record identifier, data source, and that note about the inclusion scope—only national-level projects are included, and the name, institution, and career history must exactly match public materials; those that do not are marked “Pending Verification” and are not counted as experts. Writing the criteria directly into the record header means users can at any time judge how this record came to be.
Scroll down to the employment history and research achievements sections, also presented in tables:
▲ Figure 5 | Employment history lists institution and position by time; research achievements are divided into academic contributions and projects led
What is worth noting is its restraint. There are no scores, no rankings, and no “match rate” or similar indicators in the interface—only fields that can be verified against external materials, such as time, institution, position, major, and result. This design choice is itself an attitude: only what can be verified gets written in; judgment is left to people.
05 Boundaries
① This article discusses the feature’s mechanism and does not constitute a promise of results. The completeness of a record depends on the availability of public materials; the inclusion scope is limited, and this does not mean a complete profile can be provided for every researcher.
② Structuring does not equal evaluation. A record is an organization and presentation of public materials and does not constitute an evaluation, recommendation, or endorsement of any person’s ability; judgments involving specific individuals must be independently verified by the user, who bears the responsibility.
③ Only public materials are used. No personal privacy data is involved; when used for formal purposes such as review, recruitment, or due diligence, the current rules and procedures of the relevant organization shall prevail.