{"generated_at":"2026-11-12T18:35:19.321Z","source":{"name":"Hacker News public API","url":"https://github.com/HackerNews/API","attribution":"Stories and comments are public Hacker News submissions by their respective authors. HackerHub is an independent reader, not affiliated with Y Combinator.","content_is_untrusted":true},"story_id":50456791,"hn_url":"https://news.ycombinator.com/item?id=50456791","total_comments":156,"returned":20,"scope":"Up to the requested number of top-level comments in Hacker News ranked order. Deleted comments are omitted. Nested replies are not fetched; this is not the complete thread.","comments":[{"id":50465633,"author":"muzani","created_at":"2026-11-12T11:15:59.000Z","text":"This appears the first decision AI that actually handles audio, video, and PDF, and I'm surprised this isn't highlighted everywhere.\nMarketing seems to have dropped the ball on that. I doubt anyone's picking it for the +2% higher points, but being able to answer based on a PDF or call is groundbreaking.\nJev is still faster and cheaper for many of the common text uses.","parent_id":50456791,"replies_count":2,"hn_url":"https://news.ycombinator.com/item?id=50465633"},{"id":50458144,"author":"eminence32","created_at":"2026-11-11T20:22:34.000Z","text":"I'm sort of surprised it took Jev for everyone to realize \"oh yeah, this is a useful thing that we should have in our product portfolio\".\nBut maybe this is one of those things that's only obvious in retrospect? I'm sure people were trying something similar years ago with grammar-constrained outputs (for example to only output a simple json structure), but you can't get probabilities out of the model that way. Which makes me wonder: how do these models calculate probabilities?","parent_id":50456791,"replies_count":15,"hn_url":"https://news.ycombinator.com/item?id=50458144"},{"id":50463596,"author":"v18a","created_at":"2026-11-12T06:02:37.000Z","text":"A couple of weeks back, someone had posted this on HN - an interesting note about the problems with asking LLMs for confidence: https://rumple.github.io/blog/2026/10/20/asking-llm-for-conf...","parent_id":50456791,"replies_count":2,"hn_url":"https://news.ycombinator.com/item?id=50463596"},{"id":50459215,"author":"kodoman","created_at":"2026-11-11T21:54:08.000Z","text":"I have still not used a decision model, How are people using them at the moment? is any one using them in a coding harness like Pi?","parent_id":50456791,"replies_count":16,"hn_url":"https://news.ycombinator.com/item?id=50459215"},{"id":50456959,"author":"waximabbax","created_at":"2026-11-11T18:47:29.000Z","text":"Its bit more expensive than other other Decision Models in the same lane.\nHowever, it can accept Audio and Video as input as well and for those specific cases it can be extremely useful.","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50456959"},{"id":50460195,"author":"2001zhaozhao","created_at":"2026-11-11T23:04:31.000Z","text":"We're about at the point where the AI is now pretty insanely cheap. Passing 100K tokens of context then asking hundreds of questions of 100 tokens each would cost about as much as a single Google search API call.\nOut of the hundreds of questions, most of them could even be different variations of the same question to capture more nuance than what is normally possible in a single multiple-choice.","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50460195"},{"id":50467117,"author":"vintermann","created_at":"2026-11-12T13:54:35.000Z","text":"Hm, it looks like they use Jev terminology (\"noul\") if not outright Jev syntax, but the word \"Jev\" doesn't seem to be mentioned at all. Bit of a dick move to a much smaller competitor to not acknowledge the influence.","parent_id":50456791,"replies_count":0,"hn_url":"https://news.ycombinator.com/item?id=50467117"},{"id":50458858,"author":"theanonymousone","created_at":"2026-11-11T21:20:41.000Z","text":"The best decision model I have used (besides Jev itself) is respan/span-02, which is based on (finetuned from?) Gemma4 26A4. That makes me optimistic for this release, despite the high cost.","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50458858"},{"id":50463197,"author":"flockonus","created_at":"2026-11-12T04:53:07.000Z","text":"At 0.10/M this feels very overpriced vs Liquid's 0.02/M, which on my own tests (text input) is very capable.\nWish they provided more comprehensive inputs to the model's capabilities and comparisons.","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50463197"},{"id":50459437,"author":"nico","created_at":"2026-11-11T22:12:20.000Z","text":"For anyone wanting something local, private, faster, which can also run and train on CPU, checkout https://http://jeffyclassify.com/\nIt comes with a bunch of pre trained classifiers, and you can use the tooling to train your own\nThere are also integrations with workflow tools like n8n","parent_id":50456791,"replies_count":0,"hn_url":"https://news.ycombinator.com/item?id=50459437"},{"id":50458537,"author":"abhikul0","created_at":"2026-11-11T20:55:08.000Z","text":"There’s also this guide by google to use embedding gemma 2 as a decision maker. Wonder how these compare.\nhttps://developers.google.com/edge/mediapipe/solutions/decis...","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50458537"},{"id":50467052,"author":"gtoubassi","created_at":"2026-11-12T13:49:28.000Z","text":"Can anyone report real world latency? their reported p50 seems strong but no better than I think what I've seen with typesafe JEV (which in our experience is a bit faster than openai decisions api)","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50467052"},{"id":50461867,"author":"Topfi","created_at":"2026-11-12T01:35:54.000Z","text":"Audio, Video and File input? That is very appealing and I can imagine a few very unique ways to leverage this if it works well across all modalities.\nFor what it's worth, ran my quick and dirty evals with it and got in many usage scenarios a 4x lower brier over Jev (2x - 6x depending on task), which is an immense improvement and the first time a model so significantly has outperformed Jev in my experience. p50 was also over 100ms lower though all this cost twice as much across the entire eval end-to-end.\nDare I say, this is a very impressive release by the Gemini team. While testing GPT-6 Luna lead to an honestly very pleasant interaction with OpenAI staff on here, it did perform very uncompetitively so seeing a great first effort in basic text classification performance with broad multimodality, that is great to see. Course, wouldn't be surprised if Google had such a more broadly trained classifier for years as part of internal usage for platform features which they just now commercialised, but hey, I'll take it.","parent_id":50456791,"replies_count":0,"hn_url":"https://news.ycombinator.com/item?id=50461867"},{"id":50465705,"author":"amelius","created_at":"2026-11-12T11:26:43.000Z","text":"Can't LLMs just \"degrade\" into decision APIs whenever possible/necessary?\nWhy does this need to be a separate thing?","parent_id":50456791,"replies_count":2,"hn_url":"https://news.ycombinator.com/item?id=50465705"},{"id":50458289,"author":"kaycebasques","created_at":"2026-11-11T20:33:39.000Z","text":"API docs: https://aistudio.google.com/docs/decisions-api","parent_id":50456791,"replies_count":0,"hn_url":"https://news.ycombinator.com/item?id=50458289"},{"id":50459043,"author":"wasabi991011","created_at":"2026-11-11T21:36:18.000Z","text":"Surprised how close all these models are on the benchmarks","parent_id":50456791,"replies_count":2,"hn_url":"https://news.ycombinator.com/item?id=50459043"},{"id":50458587,"author":"vesterde","created_at":"2026-11-11T20:59:18.000Z","text":"Anyone hear anything around when it might land in Vertex? As far as I can tell it's just in AI Studio for now.","parent_id":50456791,"replies_count":0,"hn_url":"https://news.ycombinator.com/item?id=50458587"},{"id":50460473,"author":"low_tech_punk","created_at":"2026-11-11T23:29:02.000Z","text":"Google shipped a \"Decision Maker\" API a while ago, see https://developers.google.com/edge/mediapipe/solutions/decis...\nApparently the two teams aren't talking to each other.","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50460473"},{"id":50458930,"author":"holografix","created_at":"2026-11-11T21:27:46.000Z","text":"Google ensuring “share of wallet” from their existing customer base with a solid but not great “we do it too” product, priced to capitalise on risk and change aversion.\nStandard practice by the big G.","parent_id":50456791,"replies_count":3,"hn_url":"https://news.ycombinator.com/item?id=50458930"},{"id":50461434,"author":"tomaskafka","created_at":"2026-11-12T00:52:25.000Z","text":"Opaque classifier as a service, for when you want to discriminate people by race and social status assumed from name, but want some other entity to be responsible for plausible deniability of the process.\nHey Jev, what mortgage rate multiplier should a person named James Smith get?","parent_id":50456791,"replies_count":1,"hn_url":"https://news.ycombinator.com/item?id=50461434"}]}