Positioning
Not a news account. Not a shitposter. Not macro commentary for its own sake. The account should feel like someone who reads launches, repos, tools, and workflows through one filter: what actually matters for builders and operators?
EKcheungAI
This should not read like a dry AI text wall. It should feel like a builder/operator signal desk: timely when useful, proof-led by default, and backed by real links, screenshots, repo pages, launch pages, charts, or context images when the post needs them.
New rule
If a post needs outside proof, it should include it. If we cannot attach a useful link, screenshot, chart, launch page, repo, or context image, we should usually not post it yet. Insight is good. Insight with receipts compounds trust.
Daily cadence
3
1 source-led post, 1 repo/workflow/proof post, 1 Threads-native relationship post.
Evidence floor
80%
Most posts should carry a real link, screenshot, repo page, chart, or public-image context.
Text-only
Low
Pure text should be the exception, not the default visual language.
Primary personality
Operator
Builder-native, useful, selective, and backed by proof more than performance.
Day-to-day posting system
This is the actual daily operating model — not a vague content pillar page.
Source-led operator take
Use when there is a real launch, product change, benchmark, company move, or thread worth reframing.
Needs: official link, screenshot, chart, launch page, or quote source
Goal: explain what actually matters beyond the headline
Cadence: 1 per day max unless there is exceptional signal
Repo / skill / proof-of-use post
Use when we find something a builder can use tonight — repo, skill, toolchain, workflow, script, or test setup.
Needs: GitHub link or real product link
Goal: show usefulness, not just novelty
Cadence: 1 daily or at least 4–5 per week
Workflow / operator lesson
Use when we have a real judgment, mistake, system rule, or shipping lesson that came from actual work.
Needs: artifact, screenshot, or a clearly-earned operational insight
Goal: turn process into signal
Cadence: 3–4 times per week
Threads-native relationship post
Use Traditional Chinese and sound like a thoughtful operator friend, not translated X content.
Needs: softer caption, one sharp observation, often a question
Goal: build trust and conversation, not just impressions
Cadence: daily
What we should avoid
- text-only hot takes with no source or image support when the topic clearly needs receipts
- generic “AI is changing everything” commentary
- news aggregation with no operator frame
- repo posts without the actual GitHub link visible
- overdesigned cards that feel like aesthetic filler instead of proof
Non-negotiable post gate
- What is the object? repo, launch, chart, screenshot, workflow, image, or artifact
- What is the angle? why this matters for a builder/operator
- What is the proof? link, image, source, or visible artifact
- What is the format? X take, Threads note, repo share, reply, or quote post
- If none of these are clear: do not post yet
Exact format templates
This is the format we should repeatedly use, so the account becomes legible and predictable in a good way.
X format 01
source-led operator take
News / launch / release post
Line 1: one sharp reframing hook
Line 2–3: what most people are reading wrong
Line 4: the actual builder/operator implication
Final line: source link or “link below”
Must include
official launch page, screenshot, benchmark, chart, or a public reference image. No dry wall unless the take is unusually strong.
X format 02
repo / skill share
Useful repo / skill post
Line 1: why this repo is worth attention
Line 2: what it actually does
Line 3: who should use it
Line 4: GitHub link directly in post or first reply
Must include
GitHub screenshot, repo page crop, or Gemini concept card paired with the real repo link.
X format 03
operator lesson
Workflow / lesson / system rule
Line 1: lesson or rule in one sentence
Line 2–3: where it showed up in real work
Line 4: the decision or takeaway
Optional: one artifact screenshot
Best support
terminal crop, screenshot of artifact, task board, preview page, or repo diff if it strengthens credibility.
Threads format 01
繁中 / relationship-building
Conversation-native Traditional Chinese post
Line 1: natural opening, not a headline
Line 2–4: one observation + one consequence
Last line: a low-friction question sometimes
Tone: human, thoughtful, slightly softer than X
Best support
context image, screenshot, natural photo, event image, or soft branded card. Caption should do most of the work.
Reply format
X / Threads
Comment / quote-reply format
Line 1: agree, disagree, or sharpen
Line 2: add one useful angle
Line 3: if helpful, point to a better signal or object
Rule
Do not reply with empty praise. Add one sharper frame, one tradeoff, or one missing signal.
New day-to-day post set
These are the kinds of posts we should actually publish — with the support object explicitly named.
X post 01
source-led operator take
Most people will read this as a feature update. I think it matters more as a workflow capture move.
If this changes default behavior, the story is distribution — not just capability.
If it does not, this is just another impressive demo cycle.
That is the part I care about most now.
Back it up with
official launch page + screenshot of the workflow or launch UI.
Threads post 01
繁中 / 關係感
我現在看 AI 產品，最先看的已經不是功能多不多，而是它有沒有改變使用習慣。
功能很容易被追平，但習慣比較難被拿走。
所以很多新聞看起來很大，我最後真正會記住的，反而是那些偷偷改變工作流的東西。
你最近每天固定會打開的一個 AI 產品是什麼？
Back it up with
product screenshot, event image, or softer Gemini/native card.
X post 02
repo share
This is the kind of repo I like sharing: not AI magic, just something a builder can actually use tonight.
Good tools reduce ambiguity, create artifacts, and leave a clean next step.
That is a much better signal than “looks cool in a demo.”
GitHub link should be visible immediately.
Back it up with
GitHub screenshot + repo URL directly in post or first reply.
Threads post 02
repo / skill 推薦
我最近越來越喜歡分享那種「今天就能拿去試」的 repo / skill。
不是那種看起來很厲害，但其實你用不到的 demo。
而是你看完會想：這個我今晚就想試一下。
這種內容比較容易慢慢累積對的人。
Back it up with
repo screenshot, repo card, or a one-screen use-case proof.
X post 03
operator lesson
A useful agent is not an infinite coworker. It is a bounded operator.
Clear scope. Clear artifact. Clear done state.
That is usually where trust starts.
Everything else is theater unless the workflow is real.
Back it up with
workflow screenshot, task artifact, preview page, or terminal crop.
Threads post 03
proof-of-work
我現在越來越相信，那種願意把過程一起打開來看的內容，比單純丟結論更有價值。
因為很多時候，真正有用的不是結果本身，而是你怎麼拆問題、怎麼做判斷、怎麼取捨。
這種內容雖然不一定最大聲，但比較容易累積信任。
Back it up with
artifact crop, checklist screenshot, preview image, or working surface.
X post 04
OpenClaw / systems
The gap between demo AI and usable AI is usually workflow design.
Good systems route work clearly, reduce ambiguity, and leave artifacts behind.
The model matters. The workflow compounds.
Back it up with
OpenClaw screenshot, route page, workflow map, or operator card + source link.
X / Threads reply
add signal
Agree with the direction, but I think the more interesting signal is distribution, not capability.
Impressive launches happen all the time.
The thing that compounds is habit formation.
Use when
a reply or quote can add a cleaner frame than the original post without sounding automated or thirsty.
Visual system for the new posting lane
Use visuals as evidence, context, or framing — not decoration.
Repo proof card
Use for
GitHub shares, skill shares, useful-tool posts
GitHub shares, skill shares, useful-tool posts
Preferred support
GitHub screenshot + Gemini/coded summary card
GitHub screenshot + Gemini/coded summary card
Rule
repo link must be visible
repo link must be visible
Source-backed news card
Use for
launches, product updates, benchmark/news reframing
launches, product updates, benchmark/news reframing
Preferred support
launch screenshot, chart, benchmark crop, or public image
launch screenshot, chart, benchmark crop, or public image
Rule
do not post generic news text without a source object
do not post generic news text without a source object
Threads-native soft visual
Use for
Traditional Chinese reflective / relational posts
Traditional Chinese reflective / relational posts
Preferred support
context image, screenshot, softer Gemini frame, or understated card
context image, screenshot, softer Gemini frame, or understated card
Rule
caption should still carry the substance
caption should still carry the substance
When Gemini is useful
- repo cards
- editorial framing cards
- quick concept exploration when we need better visual taste fast
- supporting a real source link or screenshot
When Gemini should not carry the whole post
- breaking-news claims that need real source context
- posts where the proof object is more important than the brand card
- anything that starts to feel synthetic, generic, or detached from a real reference
Reference spine
The types of links and references we should repeatedly use so the account feels grounded.
Acceptable support objects
official product / launch pages
GitHub repos and README screenshots
benchmark images, charts, dashboards, or odds snapshots
workflow screenshots, terminals, previews, and artifacts
public event/context images when they improve relevance
Evergreen examples we can post around
What the daily mix should feel like
X
sharp, source-led, evidence-backed, useful enough to save
sharp, source-led, evidence-backed, useful enough to save
Threads
native Traditional Chinese, softer, more reflective, more relational
native Traditional Chinese, softer, more reflective, more relational
Shared rule
same brain, different packaging
same brain, different packaging
What I would actually post on a normal day
Morning: 1 source-led X post tied to a real launch, thread, or shift
Afternoon: 1 repo / skill / proof-of-use post with a visible link
Evening: 1 Threads-native Traditional Chinese post with lighter but real visual support
Across the day: a few evidence-backed replies or quotes where we can add a cleaner frame