Hi HN, I’m Cyril from CTGT. Today we’re launching Mentat (https://api.ctgt.ai/v1/chat/completions), an API that gives developers deterministic control over LLM behavior, steering reasoning and removing bias on the fly, without the compute of fine-tuning or the brittleness of prompt engineering. We use feature-level intervention and graph-based verification to fix hallucinations and enforce policies.

This resonates in highly regulated industries or otherwise risky applications of AI where the fallout from incorrect or underperforming output can be significant. In financial services, using GenAI to scan for noncompliant communications can be arduous without an easy way to embed complex policies into the model. Similarly, a media outlet might want to scale AI-generated summaries of their content, but reliability and accuracy is paramount. These are both applications where Fortune 500 companies have utilized our technology to improve subpar performance from existing models, and we want to bring this capability to more people.

Here’s a quick 2-minute demo video showing the process: https://video.ctgt.ai/video/ctgt-ai-compliance-playground-cf...

Standard "guardrails" like RAG and system prompts are fundamentally probabilistic: you are essentially asking the model nicely to behave. This often fails in two ways. First, RAG solves knowledge availability but not integration. In our benchmarks, a model given context that "Lerwick is 228 miles SE of Tórshavn" failed to answer "What is 228 miles NW of Lerwick?" because it couldn't perform the spatial inversion.

Second, prompt engineering is brittle because it fights against the model's pre-training priors. For example, on the TruthfulQA benchmark, base models fail ~80% of the time because they mimic common misconceptions found on the internet (e.g. "chameleons change color for camouflage"). We found that we could literally turn up the feature for "skeptical reasoning" to make the model ignore the popular myth and output the scientific fact. This matters because for high-stakes use cases (like Finance or Pharma), "mostly safe" isn't acceptable—companies need audit-grade reliability.

Our work stems from the CS dungeon at UCSD, with years spent researching efficient and interpretable AI, trying to "open the black box" of neural networks. We realized that the industry was trying to patch model behavior from the outside (prompts/filters) when the problem was on the inside (feature activations). We knew this was important when we saw enterprises struggling to deploy basic models despite having unlimited compute, simply because they couldn't guarantee the output wouldn't violate compliance rules. I ended up leaving my research at Stanford to focus on this.

Our breakthrough came while researching the DeepSeek-R1 model. We identified the "censorship" feature vector in its latent space. Amplifying it guaranteed refusal; subtracting it instantly unlocked answers to sensitive questions. This proved the model had the knowledge but was suppressing it. We realized we could apply this same logic to hallucinations, suppressing "confabulation" features to reveal the grounded truth. While some hallucinations stem from the inherent randomness of generative models, many can be identified with the concerted activation of a feature or group of features.

Instead of filtering outputs, we intervene at the activation level during the forward pass. We identify latent feature vectors (v) associated with specific behaviors (bias, misconception) and mathematically modify the hidden state (h):

  h_prime = h - alpha * (h @ v) * v

This arithmetic operation lets us "edit" behavior deterministically with negligible overhead (<10ms on R1). For factual claims, we combine this with a graph verification pipeline (which works on closed weight models). We check semantic entropy (is the model babbling?) and cross-reference claims against a dynamic knowledge graph to catch subtle relational hallucinations that vector search misses.

On GPT-OSS-120b, this approach improved TruthfulQA accuracy from 21% to 70% by suppressing misconception features. We also improved the performance of this model to frontier levels on HaluEval-QA, where we reached 96.5% accuracy, solving the spatial reasoning failures where the baseline failed. It also handles noisy inputs, inferring "David Icke" from the typo "David Of me" where base models gave up. Full benchmarks at https://ctgt.ai/benchmarks.

Most startups in this space are observability tools that tell you only after the model failed. Or they are RAG pipelines that stuff context into the window. Mentat is an infrastructure layer that modifies the model's processing during inference. We fix the reasoning, not just the context. For example, that’s how our system was able to enforce that if A is SE of B, then B is NW of A.

We believe that our policy engine is a superior control mechanism to RAG or prompting. If you’re frustrated with current guardrails, we’d love it if you would stress-test our API!

API: Our endpoint is drop-in compatible with OpenAI’s /v1/chat/completions: https://docs.ctgt.ai/api-reference/endpoint/chat-completions

Playground: We’ve built an "Arena" view to run side-by-side comparisons of an Ungoverned vs. Governed model to visualize the intervention delta in real-time. No signup is required: https://playground.ctgt.ai/

We’d love to hear your feedback on the approach and see what edge cases you can find that break standard models. We will be in the comments all day. All feedback welcome!

Show HN: Thinkmoon.ai – Build your own Alpha Arena, trade crypto with AI Agents

The American West's most iconic tree (Ponderosa pine) is disappearing

How to Create a Design System Optimized for AI Coding

Show HN: ZON-TS 50–65% fewer LLM tokens, zero parse overhead

As humanoid robots enter mainstream security pros flag risk of botnets on legs

Hollywood-hungry Gulf states bankroll Paramount's Warner Bros bid

Show HN: Made an iPad app to sync handwritten notes to Notion

Show HN: CapSummarize – Summarize videos and generate thumbnails and clips

Show HN: I built a system to collect feedback more like force it

React2Shell serves as good reminder why JavaScript is no fun

Linux 6.19's Hung Task and System Lockup Detectors Can Provide Greater Insight

How Russia's Largest Private University Linked to $25M Essay Mill

Tempo's Testnet Is Live

The Married Scientists Torn Apart by a Covid Bioweapon Theory

Will the digital control grid inevitably fail, or is it here?

Why a College Fighting for Survival Is Slashing Econ and Physics Majors

Two Githubs, One Laptop

Interview with Calvin Rose, the Janet Creator

Meta Avocado

Barnum's Law of CEOs

Launch HN: Mentat (YC S16) – Controlling LLMs with Runtime Intervention

Nobel Winner Machado's Briefing Delayed as Oslo Arrival Unclear

Iain Douglas-Hamilton, Elephant Expert and Protector, Dies at 83

PyCharm 2025.3 – Unified IDE, Jupyter notebooks in remote dev, uv as default

I think jj-vcs is worth your time

Show HN: MPL – A Python DSL that transpiles logic to Pine Script

The AI Price Hike

The open hardware water cooling controller project

Datadog built a low-latency, multi-tenant data replication platform

Odies – AI Companions living on your screen