Harnesses often get dismissed as just scaffolding, just prompt engineering, and not real research. But that couldn't be farther from the truth. The same model weights that score 30% on ARC-AGI score 95% with a better harness.
So we gathered a group of researchers and founders working at the frontier to do a deep dive into the state of harnesses.
We cover how we got to this point, the case for making your harness as expressive as possible, and what YC learned building an agent for every employee in the company.
00:00 - @FrancoisChauba1: Why harnesses matter
04:27 - Building an auto-researcher by accident
07:13 - A five minute history of harnesses
13:56 - Self-improving harnesses
18:35 - @sethkarten: Prime Agent, a self-improving RLM harness
21:50 - Context as an L1, L2, L3 cache
24:51 - From Turing machine to von Neumann computer
28:33 - Messaging between agents
30:04 - ARC-AGI results
33:09 - Emulator Bench and GPU kernels
37:30 - @JonSaadFalcon: OpenJarvis, personal AI on personal devices
38:26 - How far behind are local models
39:21 - The five primitives of a personal AI stack
42:47 - Letting cloud models optimize your local stack
43:53 - 800x cheaper than the cloud
45:58 - @josh__france and @jbellregan: QM, YC's agent harness for work
47:29 - A history of YC's internal agents
49:24 - OpenClaw and a fleet of 50 agents
51:04 - Pulling the brain out of the sandbox
54:43 - Letting the agent choose its own sandbox and model
57:16 - The grind tool: budgets on goals
58:50 - Agents don't understand social context
This is not content, it's reality. There are people like this who just come to meetings to speak corporate jargons everytime but with no result or action to back it up.
1/5 We pretrained Omnii, a multimodal language model on DNA. Then, we asked it to design a cancer vaccine
In new research, we show how Omnii learned to turn a tumor DNA sequence into a personalized mRNA vaccine design.
Read more here: https://t.co/0Uw5yLn8pd
With AlphaFold we mapped the protein universe - now with AlphaGenome Atlas we’re charting the human genome. It can predict the impact of all 9 billion possible single-letter DNA variants, helping scientists better understand disease. Freely available for academic research: https://t.co/Gsy6lW3z6O
Not o my food industry, but sports. Will love to see how it impacts running shoes purchase for example.
GLP-1s is the single biggest medical revolution in our timeline.
GLP-1s are eating the food industry.
After households start GLP-1s, grocery spending falls 5.3%, fast-food spending 8%, and savory snacks 10%.
Among high-income households, grocery spend drops 8.2%.
Big Food has a GLP-1 problem.
Introducing Ato, your grandparent’s next computer.
Over the past year, more than 2,500 older adults around the globe have used Ato in their daily lives through our beta program, and today we're taking our next step.
Ato helps seniors stay independent on their own terms. It gives families a simple way to stay close and the peace of mind that comes from knowing how they’re doing, without cameras or tracking.
The new Ato lives in the home, becomes part of their routine, and keeps families connected, all through a simple realtime voice interface.
Millions of families are trying to help someone they love stay independent while coordinating care across relatives, caregivers, and healthcare providers. We’re building Ato into the coordination layer that brings all of them together.
Tempus trained a foundation model on 1.67M real cancer patients, fusing clinical trajectories with DNA, RNA, and H&E pathology into one patient-state embedding. Overall-survival AUC jumped 0.563 to 0.774, and treatment-benefit ranking improved in 9 of 11 cohorts.
We have argued for years that multimodal models beat curated features in oncology. This is strong evidence at real scale.
https://t.co/PBoskrC5jg
Introducing GlucoFM, a lightweight, self-supervised continuous glucose monitoring foundation model that separates metabolic baselines from transient spikes, producing transferable representations and setting new performance standards across diverse metabolic prediction tasks, such as diabetes risk assessment, insulin resistance, and post-prandial glycemic response.
Learn more →https://t.co/PA7wpFz2xo
“I’m very determined. If I decide that something is worth doing, then I’ll put my heart and soul to it. The whole ground can be against me, but if I know it is right, I’ll do it.”
- Lee Kuan Yew
El ministro Puente citaba ayer noticias obsoletas de 2024 y 2025. Ryanair actualiza trimestralmente sus datos, incluidas las entregas de aviones, información disponible en nuestra página web corporativa.
El último de los 210 Boeing 737-8200 se incorporó a la flota de Ryanair en febrero de 2026. Ninguno se destinó a la España regional, cuyos aeropuertos siguen perdiendo competitividad.
Todas las aerolíneas integradas en la Asociación de Líneas Aéreas de España (ALA) y numerosas regiones han reclamado una reducción de las tasas aeroportuarias. En Ryanair, estaremos encantados de destinar a las regiones españolas parte de los 300 nuevos aviones que recibiremos en los próximos años, siempre que sus aeropuertos vuelvan a ser competitivos y capaces de atraer crecimiento.
Chao pescao.
[1/2] At #NDR2026, Prime Minister Lawrence Wong shared how the Government will better support families moving forward. This includes enhanced childcare leave, financial, preschool and housing support.