Shared memory graph for Claude and ChatGPT, over MCP

SK

Ski

SKI is a on-device voice coding application, which can be used with any agents that supports skill, such as Claude Code, Codex or Hermes. It transcribes your voice (which you can optionally review and edit) and send it to the connected agent. The agent then completes the task, and uses the skill to talk about the updates of the project or a summary or a query using voice. This runs completely on-device. Free. No subscription. Available on both Mac (notch and pill) and Windows (pill widget) It can also be sent to meetings with the intelligence of the connected project to participate actively in the meeting. This is a paid feature (as it runs on the cloud), powered by agentcall and is optional. All on device functions are free with no limits on usage.

DD

Distilling DeepSeek into GPT-OSS doesn't transfer censorship. Try it

We recently used DeepSeek V4 Flash as a teacher for finance tasks with GPT-OSS-120B. Distillation works well on this problem. At a constrained 8k token budget, our self-distilled 120B scores 83.61% on FinanceReasoning, above Kimi K3 (81.93%) and Inkling (65.13%). We released the 20B open weights. With V4 as the teacher though, we realized it would be timely to measure if the censorship characteristic of it transferred to the distilled version of the base model. tl;dr it didn't, the teacher answered politically sensitive questions 7 SDs differently than expected, but the distilled model's behavior remained the same as its American base. You can try a couple queries yourself with no auth here: http://playground.ctgt.ai/ I will now dive in to the motivation, methodology and detailed results for those interested. The hard part of measuring this phenomena is isolating whether a model is reluctant to talk about sensitive things generally vs. a particular country's sensitive things. So we made 152 matched pairs where one prompt asked about a Chinese concept, and the other asked about a non-Chinese version of that concept. For example, the Great Leap Forward vs. the Holodomor. These were scored 0-100 by four LLM judges (Grok 4.20, Gemini 3.5 Flash, GPT-5 mini, Claude Sonnet 4.6), validated against 96 human scores at r=0.948. OpenRouter blocked some of these so we hosted the weights ourselves. The teacher's gap on the core political set of pairs was +45.45 points, ~7 standard deviations from chance, and every distilled student was within 1 point of its base. Subliminal learning literature says this is expected when the initializations are not shared between teacher and student, which is true here. The distillation data also did not contain any China-sensitive content. The contribution here was to release the evaluation framework (LineageEval: https://github.com/CTGT-Inc/lineage-eval/) to elevate the discussion around this topic in DC and beyond. We are an interpretability lab working on high risk and regulated applications of AI, so we hear a lot of vagaries aimed at the supposed dangers of distilling Chinese models on American bases. We believe these conversations should be based on open, auditable frameworks and not feelings. We plan to test what happens with a Chinese teacher into a Chinese-lineage base like Qwen next. The distillation method was an evolution of HINT-SD where we inject a hint at the specific point the model makes a mistake in its reasoning. Then we train on the corrected continuation with reverse KL over the next 100 toks of the rollout. As mentioned above 120B itself was efficacious as a teacher, and we ended up shipping this version. The self-distilled 120B scores 83.61% on FinanceReasoning, above Kimi K3 (81.93%) and Inkling (65.13%). Ours finishes 98.7% of problems in budget; the larger models truncate (90.76% and 71.01%) which score as incorrect. At 100k tokens big models gain (Kimi 89.92%). So for a finance task at a constrained (perhaps more realistic) budget a 120B on one H100 at ~$0.00026/query outpaced models running 62-160x more per query. We put out the 20B finance model as open weights (64.71% to 74.79% at 8k on FinanceReasoning, 23% lower cost/query, runs on one 80GB GPU), the 120B in a playground with teacher and students side by side (a few queries, no auth), and LineageEval with all prompts, controls, rubric, and code. We are curious to hear experiences from those working with distilled Chinese models in prod, or if you have thoughts on improvements to LineageEval. https://huggingface.co/ctgt-inc/gpt-oss-20b-finance https://playground.ctgt.ai/ https://github.com/CTGT-Inc/lineage-eval/ https://www.ctgt.ai/research/distillation-censorship-transfe...

AG

AgentCouch

Hello HN! I wanted to share a small MCP + web app I built to solve a problem at work. Essentially, we’ve been doing great when working with agents 1:1, but collaboration has been a struggle. E.g. when I have to collaborate with a teammate and share context or handoff work usually what I would do is create a markdown with my agent, paste in slack, then my teammate will copy and paste it to their agent and vice-versa. To make our lives easier, I built AgentCouch, a “messaging app” for agents that lets our agents chat and hand off work directly in shared rooms. Humans can also watch the conversation, jump in and steer it from their laptop or phone. I was aiming to solve the following issues: 1. Easily handing off context between agents by allowing them to chat directly. 2. Making sure handing off doesn’t miss important context and allowing agents to ask follow-ups. 3. Since we are also a distributed team, allowing my colleagues and their agents to talk to my agent and get their questions answered instead of waiting for me to wake up. At it's core it's an MCP with a few more endpoints and instructions(and a web UI) that allows the agents to watch the rooms. Needless to say, you have to trust the person you are sharing a room with since their agent can ask your agent anything. If you experience these issues with your team, check it out and please share some feedback if you do! Thanks!

US

Usetokenless

Hi HN, Rohit here from Tokenless (https://usetokenless.com/), which I’m building alongside co-founders Andrew and Kev. We’re building an API gateway which routes agent traffic dynamically turn-by-turn between different models to save on AI spend. The cost of AI tokens is top-of-mind for many. Companies like Uber and Salesforce have been complaining about blowing their yearly AI spend faster than expected. Frontier models are amazing for dev work, but are so expensive. Open-source models are cheap and rapidly improving, closing the gap with frontier models, but aren’t quite there yet. Tokenless gets you the best of both worlds–routing harder turns to smarter models only when needed, which keeps costs low. Before Tokenless, I was doing a PhD at Princeton. While using coding/other agents, I constantly agonized over model choice, to make sure my AI spend was going as far as possible on my academic Cursor account. At the same time, I was doing LLM research, and a small technique I developed while in recovery from NeurIPS submission season seemed to hit SOTA pretty fast. I was surprised that such simple ideas could do routing well. We’ve been able to develop a version of the router that matches the performance of Claude Fable 5 at half the cost. The blog post on our website explores the technical details on how we did this (https://usetokenless.com/blog/building-tokenless/). Highlights: - Our approach queries multiple models at once and uses their progress to make decisions (this technique is novel AFAIK, let us know if you know anyone else doing this). - Switching models doesn’t destroy the cache if the routing algorithm is aware of when the cache is hot/cold. To come: - Adding Kimi K3, all other GPT efforts and more to the router Go ahead and sign up on usetokenless.com and try using Tokenless with your agent, you’ll get $20 of free credit. Here’s a demo on how to use it: https://youtu.be/sjZWriclcls Tokenless provides frontier-level intelligence for cheaper, so we’d love some feedback on how it feels to use, any corner cases that the router routes incorrectly, and whether you find the routing problem interesting!

CH

CheapFoodMap

I was recently laid off after 18 years, and gave myself 100 days to build soething useful in public. CheapFoodMap is a crowdsourced map of meal under $10, excluding franchises, local good eats only. It's inspried by 거지맵 (Begger's Map) a Korean crowdsourced map students use to find cheap eats. Ocverage is heaviest in Texas, since I live in Dallas, but have 1200 meals across 15 US cities. Seed data came from Google Review, 4.2 star or higher with at least 500 reviews, and verified price under $10 per menu item. Things I would love feedback on : whether the price-freshness model makes sense, and what would make you trust the price on a site like this. How to encourage people to update prices, since inflation is making food price very frequent. https://cheapfoodmap.com Any and all suggestion will be super helpful. Thank you!