Whitney AI (@try_whitney) header
Whitney AI (@try_whitney) profile picture

General models wander the foothills; we find your summit.

Joined May 2026
Photo by Whitney AI (@try_whitney)Photo by Whitney AI (@try_whitney)
Last week, @OpenAI published an open letter, signed by over 200 companies, calling on security teams to continuously test their defenses against frontier attacks. Threats like prompt injection have evolved from single-turn injections to sophisticated multi-turn campaigns. To defend against these attacks, teams rely on red-teaming, simulating real-world adversaries to test defense readiness. We partnered with NuGuard AI to post-train a red-teaming agent that outperforms frontier models at lower cost and latency. We post-trained Qwen3.6-27B using a modified GRPO loop that attributes reward turn by turn instead of scoring the whole conversation at once. The specialized model lands 3.1x more attacks than frontier alternatives like Sonnet 5 and DeepSeek-V4-Flash. The attacks in the write-up are illustrative, published to advance defensive research. Full methodology in the comments.
2
2
8
1,230
Whitney AI retweeted
Jake (@jakekang ) and I interviewed Ryan Tsang (@rvtsang), who is building Harmonize Health - an AI-native operating system for medical clinics. Ryan spent four years at Microsoft working on Azure OpenAI's fine-tuning product, then trained models for open-source AI safety benchmarks at MLCommons. We get deep into how Harmonize Health builds and grades evals for voice agents (including a user-simulator agent that role-plays the patient calling in), why you should get to production before you build the perfect eval set, how Harmonize scrubs patient data out of transcripts, and the tension between forward-deployed engineering and scaling across multiple clinics. Hope you enjoy the third episode of The Interlude Show! Chapters: 00:00 - Move fast and break things doesn't work in healthcare 00:41 - Meet Ryan Tsang and what Harmonize is building 02:40 - Why healthcare, and the path from pre-med to CS 05:45 - Inside a primary care clinic: the work nobody sees 10:50 - Safety is the hardest technical problem 13:55 - Data plumbing vs. AI: EHRs and tribal knowledge 20:36 - Encoding a clinic's operating procedures into agents 23:35 - Inside their eval harness for voice agents 27:07 - What can we really not screw up? 30:16 - Healthcare moves at the speed of trust 38:57 - TempShot: pitching his high school app to us as VCs 49:53 - The YC all-nighter, quitting our jobs, and why compounding wins
3
2
14
2,716
Scam AI, a cybersecurity platform, ran honeypot AI agents to keep ten thousand scammers talking, wasting scammers’ time and collecting data to learn their tactics. Whitney AI partnered with Scam AI to build a synthetic data pipeline which allows them to scale the dataset. We post-trained a model to simulate scam call transcripts that are realistic enough to train detectors on. Our post-training recipe involves training a model that attempts to fool a frontier discriminator. We trained Inkling-Small and achieved a 25.8 percentage point increase in discriminator deception rate over the base model. Against a prompted Gemini 3.1 Pro baseline, our model leads by 12.5 percentage points. For more details on the methodology and our post-training recipe, check out our case study in the comments. As a safeguard against misuse, the ground truth, weights and code will not be released. Scammers were never short on scripts, but the anti-fraud detection side is starved for data. We help fight back.
1
1
3
840
Whitney AI retweeted
I built ML Serving and Inference architectures at Netflix and Microsoft for 4-5 years. A friend of mine (MLE @ Doordash) pointed out how far behind the US is on generative recommendation research as against China. As I followed the trail of research papers, I am more convinced that we need to focus on not just modeling but also applications of agents and LLMs. AI adoption is the biggest risk today. In this episode, @jakekang and I get into personalization, generative recommendations architecture, inference / serving optimizations and China vs. US on ML research. Hope you enjoy the second episode of the Interlude Podcast! Chapters: 00:00 - Intro 02:18 - How Netflix Personalizes Everything 04:23 - Model Serving vs. Inference 06:19 - Generative & Conversational Recommendations 08:13 - Measuring Success: Metrics & A/B Testing 11:29 - China vs. the US: Open Source & the Road Ahead 17:00 - Risks, Incentives & Regulation 20:17 - Serving LLMs vs. Classical ML & Observability
1
6
13
1,553
Whitney AI retweeted
What does working on agents look like in production? Started a podcast w/ @vatsalbajaj to talk about our experiences working on production ai systems. In the first episode, we talk about my experience building browser agents. We will bring on engineers/ctos to talk about their challenges of building AI in their industry. 00:00 - Intro 00:52 - 1st Hire at a YC Startup 02:41 - Building a self-recovering browser agent 04:18 - What are evaluations? 05:01 - 90% of the work is evals 09:46 - Offline vs online evals 12:06 - Harness engineering vs fine-tuning 15:18 - Owning your AI in regulated industries 17:35 - The OpenAI/Anthropic threat 20:52 - The future of browser automation
2
3
3
9
1,799
Whitney AI retweeted
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models. images.nvidia.com/pdf/Open-W…
16,556
34,453
12,079
193,819
69,448,410
Whitney AI retweeted
Replying to @nickmecklenburg
@nickmecklenburg and I chatted with @oneill_c , Co-Head of Model Training and @part_harry_ , Research Scientist from @baseten at @spc about the future of post-training. To lift the ceiling on context windows, they are researching KV cache compaction, and their unique take on continual learning (that it is a context, not a retraining problem) demonstrates their vision on the future of models. Learn more about Charles' and Harry's views on where post-training and continual learning are headed.
3
3
2
26
12,954
Exactly - and this flywheel compounds over time and becomes even more powerful when your learning loop feeds back into the model.
Every enterprise will have its own model-harness-sandbox-eval flywheel with token value per watt optimization. This is the future. Simple reason: tacit knowledge about the domain and customers and their workflows that the company uniquely understands and has built trust around.
5
459
1/ Nine years ago, I met my closest friends in my freshman dorm at Berkeley: @nickmecklenburg , @vatsalbajaj , and @jakekang 🐻 Lost, scared, and confused as freshmen, we bonded over the shared unknown of entering this new chapter together.
1
2
11
454
when you and your competitors have access to the same intelligence layer, maintaining a competitive edge is harder than ever -- and companies that haven't differentiated won't survive. lots of great insights from @rhythmrg on what it means to own your own ai in 2026 and why that matters. i had fun!
Your data is your edge, but only if your AI is built on it. Rent a generic model and so can your competitor. The companies with an edge are deploying custom models that they own and improve over time. Our co-founder @rhythmrg recently stopped by @spc to share how companies are owning their intelligence with Applied Compute.
1
6
220
We couldn't agree more. "This means the real opportunity is not in picking the best model but instead in building a learning loop on top of models where human capital and token capital compound. You can offload a task, or even a job, but you can never offload your learning. The future of the firm is the ability to compound that learning across people and AI."
4
181
Whitney AI retweeted
Today, @AnkurKJain, @jakekang, @nickmecklenburg , and I are excited to launch Whitney AI (@try_whitney) My best friends (and now co-founders) and I all quit our jobs and dove into the unknowns of building a startup. We started our journey this January @spc spent months wading through the idea maze. Our team comes from Microsoft Superintelligence, Netflix, Google X and YC startups. At Whitney AI, we see a massive missed opportunity. Today’s AI-native enterprises have squeezed everything they can out of frontier models via better prompts, better scaffolding and better context. But if every company is using the same general purpose models, where does that leave your competitive edge? Your real edge is in the assets that your competitors can’t copy: proprietary data, evaluations and production traces. And yet, most companies fail to harness them. Whitney AI fixes that. We embed with your team to build self-improving custom models and evaluations on your proprietary data, so the intelligence you walk away with is owned by you and compounds over time. We help you build your AI moat. Here’s a picture of my team getting ourselves flowers after a hard day at work. When we deliver for you, the next bouquet's yours.
9
6
21
985
Hello, World!
2
7
170