Ian

@BradAT There's also a suggestion that the platonic ideal of the web is plain text web pages. That might appeal to a small sliver of the population (including yours truly!) but are the masses consuming streaming TV, Reels/Tiktok and similar complaining about "ugh this video was 90 mbs"?

@BradAT There's some truth in here but a lazy critique IMO. The author implies that ad blocking would "fix" the user experience NYT. But the heaviest assets are all first-party video, photos and a bit of js:

Huge issue that no one talks about. SPO has been a thing but no one (afaik) is solving for DPO and the bloat that is creating such a bad UX on many reputable Publishers. If anyone is working on this, I'd jump at the chance to help the cause.

florida is just wario california

bro.

I foolishly assumed the “New Yorker visits SoMa for 36 hours” trope became extinct in 2019. pathetic.gif

the new "sent from my iPhone"

Is GPT-5.2 now live in the OpenAI API?

guy writing the rails app for finland's social security portal: "fucking finally"

@adamheimlich youtube.com/watch?v=Hj0lx6…

real photo berkeleyside.org/2025/10/23/liv…

I really fucked up my conference schedule this year. ... Hey Ian, We’re attending Space Tech Expo in Bremen, Germany in November and would love to connect if you’ll be there.

@BradAT x.com/avi_lum/status…

Anyone else have a take on this?

𝟣̶𝟤̶𝟪̶ ̶𝖦̶𝖡̶ ̶𝖻̶𝖾̶𝖼̶𝖺̶𝗎̶𝗌̶𝖾̶ ̶𝗒̶𝗈̶𝗎̶'̶𝗋̶𝖾̶ ̶𝗋̶𝗎̶𝗇̶𝗇̶𝗂̶𝗇̶𝗀̶ ̶𝗅̶𝗈̶𝖼̶𝖺̶𝗅̶ ̶𝖫̶𝖫̶𝖬̶𝗌̶ 128 GB because you have four Jetbrains IDEs open

@HatesAdtech @JudSpencer @Scrilla100 Notably: #overview" target="_blank" rel="nofollow noopener">docs.prebid.org/dev-docs/modul…

@JudSpencer @Scrilla100 It’s a comin on the bid verification stuff. And I totally agree. It’s astounding that none of the anti fraud providers do anything meaningful in the client prior to the auction. More stuff for us to do!

It should also be noted, that if all DSPs want to go directly to pubs via Prebid, then pubs will invent further mechanisms in Prebid/elsewhere to allow for signal fuckery and auction pumping. The Race to The Bottom will continue to be met with fierce resistance. @HatesAdtech @AdtechGod

fortunately middle of the night in the EU (and since recovered) but 😬

@BradAT Absolutely -- with pretty reasonable accuracy for most use cases. qwenlm.github.io/blog/qwen2.5-o…

AI crawlers can read screenshots?!?

@BishPlsOk @mikeosullivan @101Programmatic Thank you for _also_ providing the overly-detailed impressions! The level of time and detail you put into this is a huge compliment, truly. Re: more granular data, to what extent, if any, would a "representative sample" of the publisher meet your need?

Condensed feedback from my overly detailed impressions — current offering seems valuable for manual human quality evals of domains to block or allow them. It’s more limited for granular impression-level targeting / bid pricing without cardinality at the level of unique URL + Slot / GPID + device type, IMO. If that was available via API it would absolutely be worth paying for!

@mikeosullivan I have some initial OpenSincera questions and feedback! (Not actually looking for an in-depth Twitter response to all of this, consider it a first pass at beta user feedback and thoughts from a bit of tinkering) First, and by far the most important: is there any way to query at a more granular level than publisher domain / ID? I played around with trying to append extra query parameters to see if they'd apply filters, but I couldn't find any that worked. URL, ad slot / GPID, and device type (at least mobile / desktop distinction) would all be huge value-adds. Second: A2CR, refresh time, and average ads in view -- I assume these are populated wholly by synthetic visits from scrapers; do you have any kind of technical explainer on how the synthetic behavior is simmed? Any plans on getting into using measured data for scroll depth / scroll velocity / etc. from real user sessions to layer more reflective-of-actual-traffic-behavior accuracy onto those metrics, assuming you don't want to get into full deterministic measurement via tag-on-page integrations? What's the starting point for measuring Ad Refresh Time--total user dwell time from initial script loat, time since ad render, or viewable ad time? Methodology says "Sincera's Synthetic User will wait on a given URL and count the seconds between refresh events - a costly, but highly accurate way of determining ad refresh.", but this is a bit ambiguous -- is this done for each individual ad slot, making sure to scroll each slot into view? Is it aggregate data for all ad slots that refreshed? Is there any distinction for measuring ad slots that refresh while out of view, or that operate purely on dwell thresholds, not view time thresholds? Any metrics for whether refresh is uncapped, and/or continues while an ad slot is viewable but a user is clearly AFK / there have been no page-level user activity signals (and there's no currently-playing video media on the page)? Any data on how many ad slots are refresh-enabled VS not refresh-enabled? Is A2CR calculated for all content on on the page (I.E. HxW for larger of all ad slots containers, sum them, subtract from full page HxW?), or as a blended average of snapshots for A2CR within the observable viewport across a synthetic user session? Any distinction between above-the-fold A2CR, mid-page, bottom-of-page? (Feedback here -- slot-level A2CR while a given ad slot is in view would be valuable!) I see in the docs that A2CR uses the larger of ad slot size and ad creative size -- if an ad slot is never filled with a rendered ad (and never collapsed), is the full size of the container still fully counted for A2CR? For "average ads in view" metrics -- do you have a list of ad servers / creative vendors you support? I.E. basic GAM slots are easy, native or out of page (EYEBLASTER) could be trickier; do you use the MRC large-ads rule (30% of pixels visible rather than 50% for ads larger than the 242000 pixel threshold). Any distinction between sticky and non-sticky ads? I see the total_unique_gpids count; is there any way to break that down by unique gpids seen within a certain time threshold, any way to account for how many might be long-tail or retired / no longer in active circulation? Any way to query for WHICH publishers trip the webrisk_flagged_publishers and/or adult_domains counts? (I don't see adult_domains in the publisher response when querying directly for one the label would DEFINITELY be true of) For the ecosystem endpoint, pbjs_ad_unit_media_types is returning null. adsystem_ecosystem is present in the example response but not the live one. How are publisher slugs used? Also, for the Business Insider example slug, the docs show "open.sincera.io/publisher/busi…" but the live slug value is just "business-insider", so I'm curious on both what the purpose of the field is if it's not queryable, and how it's structured. What viewport size(s) are used for the synthetic scrapers? Single global size, or various? Any data on how frequently mobile slot sizes render into desktop units due to multi-sizing? The UI dashboard has percentile rankings for the key signals -- any way those rankings could make it into the API, whether pub-specific or the global ecosystem averages? (And on this note -- it would be VERY interesting to not just see averages, but a distribution of how many publishers fall where on the various bell curves, and potentially how many standard deviation(s) above / below an outlier is!) Small typo in the info mouseover for key signals (no apostrophe on "publisher's," screenshot attached) And last: would you name and shame the site(s) responsible for 75.2% average A2CR / 7.4 average ads in view? I can write a script to (within the rate limits, lol) pull down all the pub data to check, but I'm curious if you know off the top of your head!

for the heathens: en.wikipedia.org/wiki/Wow!_sign…

wasn't expecting to find that there

one more hit out of this one