Quick update -
We’ve been letting 100 people off the waitlist every day, but signups are still growing faster than that.
In the meantime, we’ve been bringing our top users over to the Trajectory office in SF to meet the team.
We’re working hard to speed things up so you can start training on the Trajectory platform today!
we need more whimsical interfaces for eval results!!
so excited that @pauline_makes is playing around w @trajectorylabs + built this tiny classroom to visualize her results 🙈
We have made important strides towards solving two manual bottlenecks to continual learning: reward, and envrionments.
Continual grading: Rather than needing to manually label thousands of traces, we present a metagrader - a model that can learn off of a subset of representative reward.
Group Size 1 (GS1): Instead of methods like GRPO that require groups and therefore offline environments, we explore methods were you only rollout once, proposing a stabilized algorithm GS1, where the environment becomes the product itself.
We test both of these methods on frontier, economically valuable benchmarks like PRBench, Harvey LAB, and AutomationBench and achieve frontier results.
View full blog posts below, these two pieces of our research are in private beta on the Trajectory platform
[1/3] turns out you can teach a 35b model to weight-space program better than opus o_0
earlier this week, @rikiyatakehi and @jayden_teoh_ showed that astra can manually write neural networks to solve atari games, one weight at a time. they call this weight-space programming.
on the @trajectorylabs platform, @o__jerry__o and I post-trained 27b and 35b-a3b qwen models to write networks for atari games. we curated WeightProgramBench, a set of 322 simple games to teach models to manually write neural networks.
we combined 51 atari-style games with 266 games adapted from crisp-game-lib for training, holding out the 5 original minatar games (freeway, breakout, asterix, seaquest, space invaders).
on freeway, qwen-27b climbed from averaging 21.0 to 47.1 points **without ever training on freeway**, but still not quite at astra’s 63.9 points.
Models have IQ but no experience.
Frontier models show up on day one with superhuman intelligence but no expertise on the thousands of tasks your company actually does. @trajectorylabs exists to close that gap through continual learning.
@rronak_ trained SWE-1 at Windsurf, then went to DeepMind. His cofounders @MichaelElabd and @QuantumArjun worked on Gemini Robotics and multimodal models for Vision Pro. All three hit the same wall: big labs train one model for everyone, but nobody trains one for you.
The conversation covers:
— why today’s agents are “pseudo continual learning”: we are in the paper road map era, not Google Maps
— “Group Size 1”, a GRPO variant that learns from one real-world rollout instead of replaying the same task eight times, so you don’t need a simulated environment
— how Harvey trained a model 6x cheaper than frontier that took an open-source baseline up by over 4x on its own legal benchmark
— Continual Grading, a meta-grader that turns rubrics on 100 traces into reward for 10,000
— continual learning vs. recursive self-improvement, and why only one keeps humans in the loop
— why continual learning will make product-building more like tending a garden of products
Chapters
0:00 – Introduction
0:45 – Pseudo continual learning: paper maps vs Google Maps
2:48 – Do we need new architectures? Rich Sutton, David Silver, and the missing layer
4:33 – From a $100M consulting model to a product: lessons from SWE-1 at Windsurf
7:36 – Harvey: 6x cheaper than frontier, 4% to 18% on its own benchmark
8:55 – Why "Trajectory": the unit of work is a task, not a token
10:13 – The primitives of continual learning: traces, replayable environments, reward
12:56 – Group Size 1 and the meta-grader: training without an environment or a labeled reward
16:16 – The platform: train, eval and deploy in one click, and who it's for
18:58 – Rogo and intelligence density
20:14 – What's next: curiosity, and why continual learning beats RSI on alignment
24:19 – Research and product under one roof, and the product builder as gardener
27:26 – Why work at Trajectory, and the optimistic case: from 1% to 99%
Yesterday we launched the Trajectory platform to the world, the easiest and most intuitive interface to post-train models.
Here's a preview of how you can get started with training a model with just a few agent prompts!
Today we are launching the Trajectory platform, the easiest and most intuitive tool to post-train models of your own.
The first step to continual learning is empowering every individual to train, eval, and deploy models with just a few agent prompts.
The second is the frontier research that breaks down barriers to reward functions and environments.
Here, we present Group Size 1 (GS1) and Continual Graders in beta.
We believe in a world where AI gets smarter the more it's used, and Trajectory is the way to get there.
we’re bringing together some of our fave creatives next sunday afternoon of design x storytelling
lightning talks on narrative building, writing online creative tech, filmmaking, launches and more
come join <3 ft @designengclub@osmo_studio@trajectorylabs