Novoflux SenseLab — Internal Tracker

Today's signals

∅

No signals today. Add your first.

Digest template & workflow

Reference guide for the sensing role · Export from Daily signals to generate a first draft

Capture signals daily

Use the Daily signals tab throughout the week. Flag digest-worthy items as you go.

→

Generate first draft

Click "Generate first draft" above. The export pulls flagged signals into the structure below.

→

Apply human voice

Edit the draft. Add the opener, closing note, and SenseLab angles. Write like a person, not a curator.

→

Submit for approval

Send to the editorial core team for review before publication. Never publish direct from export.

Issue structure

1. Issue header

Issue number, date, title, one-line standfirst. The title should name the week's dominant theme, not just say "this week in AI."

2. Opener — from the editors ✦ human voice required

2–3 sentences. Your editorial take on the week's mood. Not a summary of what follows — your actual observation. Write in first person. This is what subscribers remember.

"Three things landed this week that seem unrelated but all point at the same underlying shift..."

3. Signal items (2–3 main items)

Each signal: source name, headline, 150–220 words of fact then interpretation, SenseLab angle (1–2 sentences connecting to the experiment catalogue or S-curve), link to original.

4. Quick takes (3 max)

One bold headline + one sentence each. Signals worth noting that don't need full treatment. If you're writing three sentences it belongs in the main signals section.

5. Closing thought ✦ human voice required

A single observation that doesn't fit neatly into a signal item. A question several items raised without answering. Write in first person. This is where the digest becomes a conversation.

"What I keep coming back to this week is the gap between what the benchmarks show and what engineers are actually saying in practice..."

Voice & tone guide

Sound like

A senior engineer who reads widely and has a specific point of view. Someone who spent the week genuinely paying attention and has something to say — not performing expertise.

"What caught my attention wasn't the benchmark number — it was how quickly the community stopped treating it as a milestone and started treating it as a floor."

Do not sound like

A journalist summarising events neutrally. A consultant hedging everything. A vendor enthusiast who thinks everything is revolutionary.

"This development may potentially have significant implications for the broader AI ecosystem going forward..."

Sentences

Short and declarative for emphasis. Longer when building an argument. Vary the rhythm deliberately — three short sentences followed by a longer one that lands the point.

"The model release was incremental. The pricing wasn't. That gap matters more than any benchmark."

Opinions

Have them. State them clearly. It is fine to say "I think this is overstated" or "this is the development we have been waiting for." The human voice requires a position.

Word counts

Opener: 80–120 words
Each signal item: 150–220 words
Quick takes: 1 sentence each
Closing note: 80–100 words
Total target: 600–900 words
If over 1,200 words — cut the weakest signal item.

Approval checklist

Opener written in first person, not a summary Each signal has a SenseLab angle connecting to experiments or S-curve No sentence starts with "This development..." Closing note says something the writer actually thinks Total word count between 600–900 All source links verified and working Sent to editorial core team — not published direct