Path: Research & verification
Goal: Ship claim–evidence, a competitor one-pager, or a verify pass with the fewest guides—verify before bookmarking.
Best for: Students, PMs, support leads, and anyone burned by a confident wrong answer.
Chat models are excellent at sounding finished. Research habits make unfinished honesty visible: Unknown stays Unknown, tiers stay honest, and live chatter does not become policy. This path is a reading order plus weekly drills—not a promise that AI “Move from chatty answers to claim–evidence habits”
This path is a guided sequence through research guides—not a diploma. Do one linked guide fully before skipping ahead; verify beats volume.
Scene tip: Put three calendar holds this week named after research guides—not “Move from chatty answers to claim–evidence habits”
Why beginners skip this
Saving the research-path link feels finished—without a claim–evidence table this week, nothing sticks.
Why this order
First learn what models invent → a one-minute verify ritual → then matrices and one-pagers so meetings open on evidence, not vibes. 2. Source tiers → stop treating all links equal. 3. Claim–evidence matrix → meeting-ready honesty. 4. Social fact-check + Grok trend briefs → handle live noise. 5. Literature map → start new topics without fake citations.
Suggested sequence
1. When AI sounds sure but wrong 2. Source tiers 3. Claim–evidence matrix 4. Fact-check social claims 5. Trend to brief (Grok) 6. Literature map fast 7. Bonus: Competitor one-pager · Grok vs search
Example
Filled prompt
Fri research slot: competitor claimed "40% cheaper."
Claim–evidence matrix row + primary price page or Unknown.
No share of the 40% until primary open.
What good output looks like
- Calendar slot titled with slug for “Move from chatty answers to claim–evidence habits ”
- Live artifact opened — not only the bookmark
- Practice timed; refuse-line written
- Pass/fail: artifact dated Asia/Shanghai
Scene: the meeting that goes well
You walk in with a one-page matrix, not a novel. Someone pushes a rumor; you point at Status = Unknown and a verify owner. That is the path working.
Example
Filled prompt
This week research path: one competitor row Monday · source-tier drill Wednesday · claim-evidence Friday.
Assign each day a guide slug. Monday uses live vendor tabs only—no guessed prices.
What good output looks like
- Live artifact opened — not only the bookmark
- Practice timed; refuse-line written
- Pass/fail: artifact dated Asia/Shanghai
- Calendar slot titled with slug for “Move from chatty answers to claim–evidence habits ”
Scene: class / homework honesty
If this is for school, AI may help you outline questions—not invent sources. Paste your course policy boundaries into the prompt. Prefer Study path when the deliverable is graded.
Save this card
Card: claim → evidence → tier → status. No brief row without a source you opened.
Before you start
Create a 25-minute calendar block titled path-research. Open the live artifact for “Move from chatty answers to claim–evidence habits you can de”—bookmark alone does not count Keep the live artifact visible the whole time. If you only have a blank box, stop and fetch the real job first
How to use this week
This week build a claim | evidence | source-tier table for a real topic along the path; mark gaps [to-verify].
Verify beats volume
One primary source opened beats ten chat citations skimmed.
Do not bookmark-only
Finish Practice now, then tick this guide on your path. Titles unread are not skills.
Close the loop
Write one line: next time I only save the path with no table, I refuse the checkmark.
Example
Filled prompt
Live artifact this week for: Move from chatty answers to claim–evidence habits you can defend in a meeting or class.
Name real files/people. No invented facts; label Unknowns; stop before send/delete.
Return working prompt + one pass/fail check I can mark in a minute.
What good output looks like
- Calendar slot titled with slug for “Move from chatty answers to claim–evidence habits ”
- Pass/fail: artifact dated Asia/Shanghai
- Live artifact opened — not only the bookmark
- Practice timed; refuse-line written [path-research#3]
Example
Filled prompt
Fri research slot: viral claim "Category X grew 40% YoY."
Source-tier + primary or Unknown before any Slack share.
No chart screenshots as proof.
Verify before bookmarking; Unknown is allowed.
What good output looks like
- Practice timed; refuse-line written
- Calendar slot titled with slug for “Move from chatty answers to claim–evidence habits ”
- Pass/fail: artifact dated Asia/Shanghai
- Live artifact opened — not only the bookmark
Red flag
Saving the research-path link with no claim–evidence table leaves nothing—build the table first.
Pitfalls
| Pitfall | Fix |
|---|---|
| Skipping verify because the brief looks neat | Open primary tabs anyway |
| Fake citations from the model | Ban invented titles; you search |
| Treating path order as optional trivia | Do Mon/Wed/Fri drills |
| Live X chatter treated as policy | Time window + confirmed vs rumored |
Practice now
Contradiction hunt: find one source that disagrees (Asia/Shanghai).
Annotation pass: highlight claims needing sources in yellow (Asia/Shanghai).
Quote check: every quote has page/URL (Asia/Shanghai).
Quote fence: untrusted clips stay in quotes until verified (Asia/Shanghai).
Stop rule: three sources in and still no primary → park the claim (Asia/Shanghai).
Source-date drill: every kept source gets a accessed-on Asia/Shanghai date (Asia/Shanghai).
Research-path drill: pick one claim you almost cited; tier the source (primary/secondary/social) and keep or cut (Asia/Shanghai).
Thin-spot drill: rewrite the Save card with real nouns from your work (repo, doc, inbox). Generic placeholders fail the drill.
Tonight for a real topic build source tiers and a claim–evidence matrix, then write a one-page conclusion.
Today card (copy) — path-research: 1. Artifact I will produce: ___ 2. Pass/fail check: ___ 3. Refuse-line if tempted to skip verify/send: ___ (date: Asia/Shanghai) 4. One never-do from this guide: ___
Extra pass: do the card on a real file today—reading alone does not count.
One sentence takeaway
Claim | evidence | source-tier table; mark gaps [to-verify].
Next
Claim–evidence matrix · Source tiers · Research with a plan · Deep research mode · Verify before you share · When AI sounds sure but wrong
路径:研究与核实
目标: 用最少篇数跑完主张–证据、竞品一页纸或核实:收藏链接之前先核实。
适合: 书签很多、核对很少的调研与竞品跟踪者。
四十五分钟内选两篇,写出完成标准。没有链接的价格写「未列出」。
为什么按这个顺序
先知道模型会编什么,再做一分钟核实仪式;然后才上矩阵与一页纸,会上才站得住。 2. 来源分级 → 别把所有链接当同等。 3. 主张—证据表 → 开会用得上的诚实。 4. 社交事实核查 + Grok 热点简报 → 对付实时噪音。 5. 文献速览图 → 开新题时不造假引用。
为什么新手跳过
收藏研究路径标题像进步;本周没有主张证据表,什么都没留下。
建议顺序
1. 听起来很有把握却是错的 2. 来源分级 3. 主张—证据表 4. 核查社交主张 5. 热点变简报(Grok) 6. 文献速览图 7. 加餐:竞品一页纸 · Grok vs 搜索引擎
实例
填好的提示词
我有 45 分钟。建议今天在本路径选哪 2 篇调研指南来完成竞品一页纸,并写出各篇的完成标准。
只用一手页面;打不开写「未列出」,禁止编价格。
边界:不编造;未知标 [?];发送前停下。
像样的输出大概长这样
- 点名 2 篇调研指南(含路径内锚点)
- 每篇各有一条可打勾的完成标准
- 整体适合 45 分钟,不灌整条研究轨
- 终点明确是「竞品一页纸」而不是无限阅读
- 语气像教练排今日课,不是书单倾销
- 未知标出;不编造负责人、日期或指标
场景:开得顺利的会
你带进门的是一页表,不是长篇。有人甩谣言;你指着 Status = Unknown 和核实负责人。这就是路径在起作用。
实例
填好的提示词
我总在收藏链接却不核实。请写一张连到本路径「事实核实」指南的一周习惯卡。
每天 ≤10 分钟;必须含:打开一手来源、标 Unknown、禁止转发未核实。
边界:不编造;未知标 [?];发送前停下。
完成检查:一分钟内能判过/不过。
像样的输出大概长这样
- 习惯卡有触发条件(如「每收藏 3 条就核实 1 条」)
- 明确链到路径里的事实核实指南
- 以周为单位可复查,不是一次性鸡汤
- 短到能截图贴桌面
- 针对「收藏不核实」这一个坏习惯,不旁生枝节
- 未知标出;不编造负责人、日期或指标
场景:课堂 / 作业诚信
如果用于学业,AI 可帮你梳问题:不能替你编来源。把课程政策边界贴进提示词。交付物要计分时,优先走学习路径。
保存这张卡
记下:本周核实习惯触发条件(如每收藏 3 条核实 1 条)。
开始前
写清今天调研问题一句话,再打开任何指南。
本周怎么用
本周按路径对一个真主题建主张|证据|来源等级表;缺证据标 [待核]。
核实胜过堆量
打开一个一手来源,胜过扫十个聊天引用。
不要只收藏
先做完「现在就练一次」,再在学习路径上打勾。只收藏不练,标题变不成技能。
闭环
写下一行:下次只收藏路径不建表,我拒绝打勾。
红旗
收藏研究路径却没有主张证据表,什么都没留下;先建表。
常见坑
| 坑 | 改法 |
|---|---|
| 收藏不核实 | 每收藏 3 条至少核实 1 条 |
| 45 分钟想读完整轨 | 只选 2 篇,写出完成标准 |
| 竞品价格凭记忆 | 没有链接就写「未列出」 |
现在就练一次
今晚就练:对一个真主题建主张|证据|来源等级表;缺证据标 [待核]。
一句话带走
主张|证据|来源等级表;缺证据标 [待核]。
下一步
← Back to home · Learn · Tools · Compare · Start