That one is a bit sus to me, because the models that do the worst on Omniscience Accuracy do the best on non-hallucination. The top model for this benchmark is "MiniCPM5-1B (Non-reasoning)" which gets a whopping 99% vs 45% for Fable 5.
I'd love to see a good hallucination benchmark, but this isn't one. There's no possibility that a 1B model hallucinates less than Fable 5.
Most of those inference providers are American, and China is actually at a disadvantage here because of export restrictions - US companies are using newer and more efficient chips.
I'm running it for the first time and this is what the thinking looks like. Opus seems highly concerned about whether or not I'm asking it to develop malware.
> This is _, not malware. Continuing the brainstorming process.
> Not malware — standard _ code. Continuing exploration.
> Not malware. Let me check front-end components for _.
Well I suppose I can't miss out on the opportunity to plug my open-source menu bar app for voice-to-text in any app! Going on three years of development, believe it or not.
When I set up a Mac I have a short list of things that I need to install. When I set up Windows I have a much longer list of things that I need to un-install. I much prefer the former.
iStat menus (or the new open source "Stats") is a brilliant use of the menu bar, but it can take a lot of space!
Then couple ordinary services that add menu bar icons that you don't even ask for (DropBox, Docker, Adobe, etc.) and you can overflow onto the notch quite quickly.
Every time I get a new Mac, I run these commands to reduce the spacing between menu bar icons. Lets you fit at least 2x the number of items in the menu bar.
This is exactly the case today. Multimodal LLMs like gpt-4o-transcribe are way better than traditional ASR, not only because of deeper understanding but because of the ability to actually prompt it with your company's specific terminology, org chart, etc.
For example, if the prompt includes that Caitlin is an accountant and Kaitlyn is an engineer, if you transcribe "Tell Kaitlyn to review my PR" it will know who you're referring to. That's something WER doesn't really capture.
BTW, I built an open-source Mac tool for using gpt-4o-transcribe with an OpenAI API key and custom prompts: https://github.com/corlinp/voibe
I created Voibe which takes a slightly different direction and uses gpt-4o-transcribe with a configurable custom prompt to achieve maximum accuracy (much better than Whisper). Requires your own OpenAI API key.
Do you mean like structured outputs? Unfortunately here the model is guided to explicitly tell you when you violate the rules and why, it can confuse it's system rules with the game rules and say you're not allowed to ask a question about copyrighted material etc.
The performance is great, but the censorship is ridiculous for me. I tried it as a backend for my game Guessix[1], but it would refuse for ridiculous reasons like "Cannot answer questions about copyrighted works like Harry Potter."
ParaFi Capital | San Diego (Hybrid) | Sr. full-stack Engineer
ParaFi is a multi‑strategy investment and technology firm focused on the digital‑asset ecosystem. Since 2018 we’ve partnered with leading institutions to deploy venture, liquid, and quantitative capital. Our edge comes from marrying rigorous research, deep technical capability, and an operator’s mindset. We ship code, contribute to protocols, and build internal systems that keep us a step ahead of the market.
We’re looking for a senior engineer who can wear multiple hats - backend, frontend, data, and DevOps - to accelerate every part of ParaFi’s technology platform. You’ll design and own production systems that ingest billions of on‑chain and market datapoints, power quantitative strategies, and surface realtime insights to the investment team. Expect a high‑trust, low‑bureaucracy environment where your technical decisions directly influence portfolio returns.
To apply, send me an email directly and mention HN in the title. Just decode this email: [email protected]
Torrey Pines area definitely looks the most different, mostly because of the growth of UCSD I'm thinking.