Michael Elabd

Co-founder @TrajectoryLabs
Ex-Research @DeepMind, @Google, @Stanford

[Michael Elabd @MichaelElabd]

Today we’re announcing the launch of @Trajectorylabs, where we’re building the platform for continual learning.

The research question motivating us is simple: can AI systems improve in response to real-world experience?

Today’s agents are episodic: they complete a task, receive feedback, and reset. In doing so, they miss valuable learning signals: retries, edits, user interventions, etc. By closing the loop between interaction data and model improvement, we believe that agents can continuously improve.

But honestly what excites me most about Trajectory is working with such talented people. Our team has already trained models that beat SOTA on customer evals, models that are operating in production today.

AI research has long aspired toward continual learning. In our work thus far, we’re already seeing early signs of this being a practical reality.

[Ronak Malde @rronak_]

Today, @MichaelElabd, @QuantumArjun, and I are excited to announce Trajectory.

We are a research lab and product company building the platform for Continual Learning.

Our platform unlocks the signal already sitting in product usage, so companies can continuously post-train large-scale agentic models that outperform the frontier. @trajectorylabs

We’ve raised $15M from @Conviction, @BessemerVP, @radicalvcfund, @jeffdean, @drfeifei and more.

We’re partnering with some of the best AI-native companies: @ClayRunHQ @Harvey, @DecagonAI, @mercor_ai, @RogoAI to power their agentic systems, some of which we are already in production with.

We’ve brought together a world class research team from DeepMind, OpenAI, Apple, Meta Superintelligence, Amazon AGI, Scale AI, and an elite product team from Stripe and Figma.

AI will never again start on day one. Every correction, every retry, every edit will make products smarter. This is Continual Learning.

[Michael Elabd @MichaelElabd]

At Trajectory, our goal is to bring continual learning to every company.

That means training on production data as it actually arrives: one task, one trajectory, often stale with respect to the current model.

Making SDPO work on long-horizon agentic tasks is a major step toward env-free RL and real online learning!

[Trajectory @trajectorylabs]

5 Days of Trajectory

🏹Day 5: Scaling SDPO to Agentic Tasks

Continual learning means you must train on data from production. But production gives you one example per task. A user makes a request once. You get one trajectory, not a batch.

However, current RL algorithms don't work that way, They need groups of tasks. By definition, that means you need some artificial environment to perform those rollouts in. But what if you don't?

SDPO is a promising route. It learns from a single trajectory, with no group required and failures still producing signal. The shape of the method matches the shape of production data.

But one fundamental problem remained. Every published SDPO work assumed fresh, on-policy rollouts. Agentic work cannot give you that. Trajectories run for an hour or more and arrive stale. On true agentic tasks, naive SDPO collapses.

We fixed it. We're the first to make SDPO work on agentic tasks.

On Mercor's APEX-Agents, with hour-long trajectories and near-zero base pass rates: 25% average reward, 5x over zero-shot. More importantly, it trains stably and the curve is still climbing.

Read more below.

[Michael Elabd @MichaelElabd]

@GoogleStartups @trajectorylabs Been amazing partnering with GCP! Thank you for all your help

[Michael Elabd @MichaelElabd]

@lialby @trajectorylabs @QuantumArjun 👀

[Michael Elabd @MichaelElabd]

Day 4 of Trajectory!!!!!!

For today, we are showcasing our vision to the world. What will the world look like? Where do product companies fit into this world? How will software change and evolve over time?

Beautiful writing by our @QuantumArjun, really crisp telling of how we think the world will evolve.

[Trajectory @trajectorylabs]

🏹 5 Days of Trajectory.

Day 4 - Why We’re Building Trajectory

AI is the most capable software ever built.

You correct it.
You teach it what you want.
However, the next session starts, and the learning is gone.

This is deeply unnatural - nothing intelligent works this way.

Today, we’re sharing the thesis behind Trajectory:
- why continual learning is the next platform shift in AI
- why the primitive governing that shift is the trajectory
- our plan to move products from being shipped to being grown: first make the intelligence layer better, faster, and cheaper; then make it shapeable; finally, make it learn

Read more below⬇️

[Trajectory @trajectorylabs]

We’re taking a quick break for the 5 days of Trajectory, but wanted to take this time to say that we’ve been named to @Redpoint’s 2026 Infrared 100 as one of the companies shaping the future of AI infrastructure.

We're so grateful for the recognition so early in our journey, and want to congratulate the other awardees as well!

[Michael Elabd @MichaelElabd]

Full piece here: https://t.co/sSl353pMnI

[Michael Elabd @MichaelElabd]

Day 3 of 5 Days of Trajectory 🏹

Continual learning is the north star 💫. We belive models should improve hourly from real production use.

But frontier-scale training makes that hard. Spinning up massive jobs across GPU nodes over and over is slow and expensive (and operationally painful for researchers lol).

So we built Continual-LoRA: many lightweight adapters training concurrently on one shared base model.

Instead of splitting one giant job across nodes, we load-balance many small jobs over a single base.

The result: 2.81x experiment throughput over single-tenant training, with no reward regression.

Proud to open-source this in SkyRL with @anyscalecompute , @NovaSkyAI as one of the first multi-LoRA RL training platforms.

Excited to see what teams build with it. If you’re thinking about continual learning, reach out.

[Trajectory @trajectorylabs]

🏹5 Days of Trajectory.

Day 3 - An Open Source Training Stack for Continual Learning

Building the platform for continual learning requires both partnering with pioneering AI companies, as we showed on Day 2 with Harvey, and working toward frontier research, which we are highlighting today.

Continual learning means models that improve hourly from real production use. But with the size of frontier models, this becomes quite difficult. A Qwen-397b would need to spin up and tear down repeatedly across six GPU nodes, and that's valuable time gone.

Our contribution is Continual LoRA (C-LoRA): many lightweight adapters running at once on one shared base model. Our insight centers on where the parallelism lives: instead of splitting one giant job across nodes, we load-balance many small jobs over a single base.

The result: 2.81x experiment throughput over single-tenant training, with no regression on rewards.

We built this together, with @anyscalecompute, @NovaSkyAI, and generous support from @GoogleCloud and @GoogleStartups. We've open-sourced on SkyRL as one of the first multi-LoRA, RL training platforms, so that every team can get to continual learning faster.

We’re very excited to see what you build, please reach out!

[Michael Elabd @MichaelElabd]

@trajectorylabs Fire work by @j316chuck @yapdianang @hersh_godse @o__jerry__o and @QuantumArjun!! Really proud of the team here 👏

[Michael Elabd @MichaelElabd]

@j316chuck @erictang000 @charlie_ruan @sumanthhegde @pcmoritz Fire analysis!!! Whats next for you to Chuck??

[Michael Elabd @MichaelElabd]

@sumanthrh @j316chuck @pcmoritz @tyler_griggs_ Thank you so much @sumanthrh the partnership has been very fruitful to us! Really excited to keep building this openly!

[Michael Elabd @MichaelElabd]

@j316chuck @trajectorylabs @harvey @o__jerry__o LOL

[Michael Elabd @MichaelElabd]

@ryancjulian Honestly pretty cracked models! Cant wait to get our hands on Nemotron 3 ultra

[Michael Elabd @MichaelElabd]

Day 2 of Trajectory!!

We partnered with Harvey to post-train NVIDIA Nemotron 3 Super on their new LAB benchmark. Results? Frontier-level legal reasoning, a fraction of the cost, and enterprise-grade sovereignty.

We are also excited to announce that Nemotron 3 Ultra is coming very soon!

(So excited for the next day of Trajectory!!!)

[Trajectory @trajectorylabs]

Welcome to Day 2. Yesterday, we showed the broader work we're doing with the pioneers of continual learning.

Today we'd like to deep dive on one: how we post-trained an open model for legal work, in partnership with @Harvey.

We've built a platform where production data is the moat. Every correction, retry, and edit becomes signal you can post-train on, and the models are plug and play: customer's can drop in their model of choice, and improve from there.

Fields like legal and finance make those demands absolute, with hard security, sovereignty, and provenance requirements. That's why we post-trained @nvidia 's open-weight Nemotron 3 Super, on Harvey's LAB benchmark.

The results, in just hours: post-trained Nemotron 3 Super approaches the closed frontier, matches GPT 5.5, lifts rubric-pass criteria +25%, all while beating the performance-vs-cost frontier. That's the power of our platform.

And this is just a glimpse towards what the future of intelligence will look like: continual learning, where products get smarter every time they're used.

Thanks to @nikogrupen, @gabepereyra, @ItsJulioPereyra, and the whole Harvey team for their collaboration on this. Much more to come soon on continually learning legal agents

[Michael Elabd @MichaelElabd]

@gefkovicz @NVIDIAAI @harvey @trajectorylabs @rronak_ @QuantumArjun Its a team effort!! The team has been executing at break neck speed. Also huge thanks to you and all of our investors for always being there for us! 🙏

[Michael Elabd @MichaelElabd]

@NVIDIAAI @harvey @trajectorylabs It has been amazing working with Nemotron @NVIDIAAI! Excited to scale these results further as well, to get to continual learning for legal!

[Michael Elabd @MichaelElabd]

@gabepereyra @rronak_ @QuantumArjun @trajectorylabs @NVIDIAAI Really excited for this partnership Gabe! Lets push the frontier of legal work!

[Michael Elabd @MichaelElabd]

@MinseokMatthew @trajectorylabs Ya i think scaling up group-free/env-free RL is one of the most interesting research directions of this year! We are actively exploring this research as well and will be sharing more soon

[Michael Elabd @MichaelElabd]

Today we’re announcing the launch of @Trajectorylabs, where we’re building the platform for continual learning.

The research question motivating us is simple: can AI systems improve in response to real-world experience?

Today’s agents are episodic: they complete a task, receive feedback, and reset. In doing so, they miss valuable learning signals: retries, edits, user interventions, etc. By closing the loop between interaction data and model improvement, we believe that agents can continuously improve.

But honestly what excites me most about Trajectory is working with such talented people. Our team has already trained models that beat SOTA on customer evals, models that are operating in production today.

AI research has long aspired toward continual learning. In our work thus far, we’re already seeing early signs of this being a practical reality.

[Ronak Malde @rronak_]

Today, @MichaelElabd, @QuantumArjun, and I are excited to announce Trajectory.

We are a research lab and product company building the platform for Continual Learning.

Our platform unlocks the signal already sitting in product usage, so companies can continuously post-train large-scale agentic models that outperform the frontier. @trajectorylabs

We’ve raised $15M from @Conviction, @BessemerVP, @radicalvcfund, @jeffdean, @drfeifei and more.

We’re partnering with some of the best AI-native companies: @ClayRunHQ @Harvey, @DecagonAI, @mercor_ai, @RogoAI to power their agentic systems, some of which we are already in production with.

We’ve brought together a world class research team from DeepMind, OpenAI, Apple, Meta Superintelligence, Amazon AGI, Scale AI, and an elite product team from Stripe and Figma.

AI will never again start on day one. Every correction, every retry, every edit will make products smarter. This is Continual Learning.

[Michael Elabd @MichaelElabd]

@ClementDelangue @rronak_ @QuantumArjun Would love to!

[Michael Elabd @MichaelElabd]

@harvey @trajectorylabs Very fruitful partnership! Really excited to see how we can move the poreto frontier together!

[Michael Elabd @MichaelElabd]

@lialby @trajectorylabs @harvey LOL maybe we should @ him

[Michael Elabd @MichaelElabd]

@jennzhaii @trajectorylabs @harvey @o__jerry__o 🐐