Daniele Vella

Co-founded CommerceClarity, where AI agents run retail catalog operations.

[Daniele Vella @_DaniVella]

@GergelyOrosz this looks like a storage fallback leaking straight into the UI... if the value means unlimited, it needs its own semantic state, not a bigger number

[Daniele Vella @_DaniVella]

@algolia @jdsportsus the numbers are strong, but i'm curious where the agent enters the flow... are these already agent-led sessions, or is the search layer being prepared for them?

[Daniele Vella @_DaniVella]

@dhh @rails @OpenAI i'd want the cost per completed task next to this... is that $11 for the whole eval, or per successful task?

[Daniele Vella @_DaniVella]

🚨 OpenAI just added Product Feeds to Ads Manager.

ChatGPT is becoming another distribution channel for product data, alongside Google, marketplaces and ecommerce storefronts.

For retailers, catalog quality and structure just became even more important.

[Daniele Vella @_DaniVella]

@oderajoseph the PR can be right while the running system is still wrong... the last check has to hit the live route, not just the new build

[Daniele Vella @_DaniVella]

@EmmerichMarcelo i've implemented WebMCP and the hard part wasn't exposing actions, it was deciding what state and permissions each one should carry... do you put policy in discovery, or only enforce it when the tool runs?

[Daniele Vella @_DaniVella]

@lucaronin small is usually easier to define by ownership than size... one context, one decision, one clear output

[Daniele Vella @_DaniVella]

@VaibhavSisinty owning the runtime is one thing, making it trustworthy on live legal work still comes down to evals, permissions and auditability

[Daniele Vella @_DaniVella]

@YTforBusiness 5x purchases sounds big, but I’d love to see how much comes from the creator versus the recommendation layer

[Daniele Vella @_DaniVella]

@dresserman the 8% volume is the real signal here, curious how much comes from range and availability versus promo intensity

[Daniele Vella @_DaniVella]

@harleyf @Shopify @OpenAI catalog sync makes setup easy, i'm curious about the ugly edge cases... when price or availability changes, does the ad update as fast as the storefront?

[Daniele Vella @_DaniVella]

@PimVendors that separation is doing more work than most PIM models admit... how do they handle a packaging change that affects only some markets?

[Daniele Vella @_DaniVella]

@andrewqu curious, which commerce-specific check changed the score most compared with the generic audit?

[Daniele Vella @_DaniVella]

@lilyraynyc did the failures happen page by page, or was there a point where the whole pattern started getting discounted?

[Daniele Vella @_DaniVella]

@agentcardhq @keyserfaty @tbpn is that amount cap global today, or can the vault also restrict merchant and category?

[Daniele Vella @_DaniVella]

@sarah_edo I wrote about my point of view here https://t.co/d7E2txG8fR

[Daniele Vella @_DaniVella]

Google is starting to score websites on how ready they are for AI agents.

We tested llms.txt + WebMCP on CommerceClarity.

What matters for ecommerce today, what doesn’t, and what I’d actually implement ↓

https://t.co/d7E2txG8fR

[Daniele Vella @_DaniVella]

@GergelyOrosz Perf Factory is the interesting bit for me too, especially if the eval loop can catch regressions before humans review the change

[Daniele Vella @_DaniVella]

@Pangolinfo_API The dangerous failures are the ones that keep the pipeline green. Do you treat semantic meaning as part of the field contract too?