Artem Vidiaev

User acquisition, application and service development. No bullshit. In love with BJJ.

[Artem Vidiaev @artem_vidiaev]

$3.5 marketing spend → $500K+ revenue in 6 months.

No clickbait. Real case study I can finally share - the account got banned, the app is dead, nothing left to protect.

Here's the full playbook:

[Artem Vidiaev @artem_vidiaev]

@ParthProductX I’d trigger the offer from behavior, not a timer: let users see the original price, then show the discount only on explicit exit or cancellation intent. Otherwise you risk training them to wait for the cheaper price.

[Artem Vidiaev @artem_vidiaev]

@YapamarMMM Ranges and decision thresholds are more useful than extra decimal places. If a small attribution swing changes the decision, the test probably isn’t robust enough yet.

[Artem Vidiaev @artem_vidiaev]

@soniaia_ That changes the job of measurement. A dashboard is not enough; teams need to explain assumptions, uncertainty, and how attributed revenue reconciles with the P&L

[Artem Vidiaev @artem_vidiaev]

@arindam___paul One extra audit: check whether new concepts are actually reaching new people, not just wearing a new hook. Creative diversity should expand the audience and the underlying angle, not only the asset count

[Artem Vidiaev @artem_vidiaev]

@_evancarroll AI statics and AI UGC are solving different problems here. Statics benefit from iteration speed; UGC still depends heavily on credibility. I’d separate the format effect from the AI effect before making the 90% call

[Artem Vidiaev @artem_vidiaev]

@ivesparrowai It's pretty funny watching people genuinely not understand what these guys are getting banned for and writing "Why?")

[Artem Vidiaev @artem_vidiaev]

@SimonHoiberg The defensible layer is shifting from features to workflow depth, proprietary context, and distribution. If a user can rebuild the core in a weekend, the product needs to own a harder problem

[Artem Vidiaev @artem_vidiaev]

@rowancheung The authority doc matters more than a bigger context window. Long projects usually drift because decisions stay implicit, not because the model forgot the raw conversation

[Artem Vidiaev @artem_vidiaev]

@kr0der The risk with minimum spend is forcing budget into weak concepts before they earn it. I'd separate concept discovery from scaling: controlled spend for clean reads, then consolidate winners once the signal is real

[Artem Vidiaev @artem_vidiaev]

@ZhuAvery24866 The key is whether the tutorial shortens time to first value, not whether users complete it. Rewards can lift tutorial completion while teaching very little about the core habit.

[Artem Vidiaev @artem_vidiaev]

@Tardammm The middle ground is a small paid test before scaling. It can show whether retention and IAP hold beyond an unusually motivated organic cohort, without pouring real budget into a leaky product.

[Artem Vidiaev @artem_vidiaev]

@AHovhannisians Active parameters are only one part of local inference economics. Total weights still affect memory and storage, while routing and memory bandwidth shape latency. The useful benchmark is tokens per second on the actual target hardware.

[Artem Vidiaev @artem_vidiaev]

@Rohthebuilder @RevenueCat Those two medians don’t compare cleanly. One is LTV per paying subscriber; the other is CPA across mixed industries and conversion events. The useful comparison is cohort CAC per payer vs. realized LTV for the same geo, channel, and window.

[Artem Vidiaev @artem_vidiaev]

@echelonnio The fastest sanity check is reconciling platform purchases against backend orders by day, not comparing ad dashboards to each other. If that gap is wrong, every ROAS and bidding decision downstream is noise.

[Artem Vidiaev @artem_vidiaev]

@DhravyaShah @supermemory Permissions are where an agent turns into an actual product. The hard part isn’t just data access - it’s keeping delegated actions narrow, revocable, and auditable without asking for approval on every run.

[Eric Seufert @eric_seufert]

OpenAI introduces conversion optimization, app install attribution, and advanced matching https://t.co/7yhgTsEetC

[Artem Vidiaev @artem_vidiaev]

@doodlestein Trust should probably be scoped by task, not model. An agent can earn autonomy on tests and refactors while still needing tight review on migrations, auth, or billing. Blanket trust is where the expensive failures start.

[Artem Vidiaev @artem_vidiaev]

@mufvza One caveat: don’t postpone monetization too long. Price is part of validation, and trial-to-paid behavior can expose a weak habit faster than retention alone. Validate the loop and willingness to pay in parallel; optimize later.

[Artem Vidiaev @artem_vidiaev]

@eastdakota New entrants may need to treat paid social less as the conversion endpoint and more as a demand layer that produces reviews, comparisons, and first-party signals agents can verify. Brand still matters; the evidence trail around it matters more.

[Artem Vidiaev @artem_vidiaev]

RevenueCat notifications are fun until your app gets product-market fit