Collinear AI’s Blog

Collinear Newsletter #11 - Notes on Frontier AI

Hi AI innovators,

Apr 30 • Soumyadeep Bakshi

10

AI's U-235 Problem

Nuclear physics solved for k_eff. What's the AGI equivalent?

Apr 23 • Jed Gresham

8

SimLab: The self-serve staging playground for real-world agents

Agents fail on real tool calls, long workflows, and messy data. SimLab lets you find those failures in simulation, not in production.

Apr 2 • Sachin

10

March 2026

Collinear Newsletter #10 - Notes on Frontier AI

Happy March from the Collinear team!

Mar 24 • Soumyadeep Bakshi and Nazneen Rajani

10

We gave Claude, Gemini and GPT, $250k, and it didn't go as you’d expect...

Introducing YC Bench: The first open-source, long-horizon benchmark with a simulation clock

Mar 5 • Muyu and Anand Kumar

7

1

December 2025

Collinear Newsletter #9 – Notes on Frontier AI

Hi AI innovators,

Dec 19, 2025 • Soumyadeep Bakshi and Nazneen Rajani

5

November 2025

RL Infrastructure for AI Agents: Why Environment-as-a-Service is the Missing Piece

Reinforcement learning for large language models is more of a systems problem than ML.

Nov 18, 2025 • Nazneen Rajani

3

1

Announcing Spider: a lightweight tool to craft post-training data recipes

TL;DR Spider is a single client interface that turns messy distillation and ablation experiments into a simple, configurable workflow.

Nov 6, 2025 • Soumyadeep Bakshi

6

Collinear Newsletter #8 – Notes on Improving AI

Hi AI innovators,

Nov 4, 2025 • Soumyadeep Bakshi

8

October 2025

The case for simulations

Unlocking model uplift through better evaluations

Oct 23, 2025 • Soumyadeep Bakshi and Grant Griffith

6

Through the Valley of Reasoning: What Small Models Teach Us About Learning

NeurIPS paper on knowledge distillation scaling laws for small foundation models

Oct 9, 2025 • Muyu, Tsach Mackey, Anand Kumar, Meghana A Rajeev, and Nazneen Rajani

1

Introducing Collinear Simulations: Steerable Personas for AI Agent Testing

TraitBasis, inspired from mech intrep, gives high-fidelity user personas for comprehensive agent testing

Oct 7, 2025 • Meghana A Rajeev, Tsach Mackey, Muyu, Anand Kumar, and Nazneen Rajani

2