Deep Insight Labs (@DeepInsightLabs)

Pinned post

Deep Insight Labs @DeepInsightLabs Oct 7

We started Seldon less than 3 months ago as an API gateway, then a compiler to help make LLM powered applications run cheaper, faster and more reliably.

We have seen a 300+% MoM user growth in that period with many interesting client conversations that have eventually led to our expanded focus on this concept of Semantic Compute: helping existing AI apps run repeated work across code, specialized models (Jev, GLiNER), and LLMs.

We’re still finding the right developer experience, full details to be revealed soon.

What have you tried to move off an LLM?

0:50

Deep Insight Labs @DeepInsightLabs 1h

It is pretty serious. I think it's not an understatement to say that serious authors have been crippled by the rampant abuse of AI, and the current process badly requires a revamp.

cyanide @cyanide1013 1h

심각하긴한가보네 선배님들 말만 들어봐도 흠
View quoted post on X

Deep Insight Labs @DeepInsightLabs 2h

Hot take 🔥

RLVR was a shortcut that has turned into a tarpit.

Many knowledge work has subjective/contextual quality measures that are usually not self-verifiable or easily measured by a rubrics.

This is why despite recent frontier models acing the most popular benchmarks out there, the real world performance still feels very inadequate.

Deep Insight Labs Verified account @DeepInsightLabs 5d

x.com/i/article/2117…
View quoted post on X

Deep Insight Labs @DeepInsightLabs 5h

How much of these is due to a saturation in the quality/realism of RL env I wonder.

Post training results from the likes of Harvey and other AI startups with domain expertise has shown that the limitations is probably more in post training and RL env.

Ramez Naam Verified account @ramez 11h

Anthropic is gaining exponentially less capability per billion dollars spent over time. It's now 8x as expensive to boost AI capabilities as it was 2 years ago, getting worse on a log scale. This includes METR report & internal models.

Credit to @alexolegimas, @sebkrier, and @michael_wiebe for inspiration.

View quoted post on X

Deep Insight Labs @DeepInsightLabs 5h

More RL env releases are upcoming. Look out for our announcements as they happen!

We are also looking to collaborate with researchers and industry practitioners on any of these areas

Program synthesis and system one models;
AI workload optimization;
RL envs in multi agents and partially observable scenarios
Inference optimization

If you are working in any of these areas, let's connect🤝

Deep Insight Labs Verified account @DeepInsightLabs 2d

SkillProbe — Can AI recognize genuine technical competence?

We're designing SkillProbe, an RL environment that trains AI agents to recognize genuine technical competence, rather than merely conduct convincing interviews.

The problem: an AI interviewer can ask 100 technical questions and still fail to identify the best engineer.

Real competence is often revealed through follow-up questions, unfamiliar problems, contradictory answers, and how candidates reason when they don't immediately know the solution.

What makes this environment interesting is the depth of real-world data informing it:

• 1M+ recorded technical interviews (~175TB)
• 1.6M+ candidate CVs and 55,000+ job descriptions
• 50,000+ curated interview questions spanning 700+ technical domains
• Structured human assessments, scoring reports, and historical hiring outcomes
SkillProbe would simulate candidates with hidden competencies, varying technical depth, and imperfect communication.
Agents must choose what to investigate, challenge unsupported claims, adapt their questioning, and reach defensible conclusions.
Crucially, rewards would be based on independently verifiable evidence of competence, not simply agreement with historical hiring decisions.

For frontier labs, this opens up a rich training ground for adaptive reasoning, active information gathering, calibrated judgment, and understanding human intent.

At Archytas AI, our vision is to turn authentic human interactions into RL environments that reward intelligence demonstrated through evidence, not eloquence.

We're looking for frontier AI labs and agent developers interested in exploring this environment, as well as researchers and collaborators working on agentic evaluation, assessment, and human simulation.

DM us or reach out at archytas-ai.com.

View quoted post on X

Deep Insight Labs

Latest posts 5