AI Endurance / Intervals.icu Integration

Hi Intervals.icu community,

We’re excited to announce that you can now connect Intervals.icu with AI Endurance.

AI Endurance: We provide AI-driven training plans tailored to runners, cyclists, and triathletes, helping you reach your performance goals with personalized, adaptive plans.

How the integration works: Once connected, your AI Endurance training plan is automatically synced to Intervals.icu. Any changes in your plan are instantly reflected on Intervals.icu, so you’re always up-to-date. You’ll be able to see your upcoming workouts for the week directly on Intervals.icu.

As a co-founder of AI Endurance, I’m happy to answer any questions. We’d love for you to give it a try and share your feedback!

Thanks,
Markus

23 Likes

Appreciate you integrating with Intervals. I had not heard of AI Endurance but I’m looking into AI plans for my 2025 season so your timing is good. I will evaluate it. The main issue I see with AI is repetitive workouts. One platform served me the same video 3 times in one week and another had almost the same workouts each week but with an increasing duration for the session.

I am a Wahoo SYSTM user and love the content. Would AI Endurance be able to serve up SYSTM workouts on a training plan if synced to Wahoo or does AI Endurance push its own workouts?

Thanks for your time

Hi James,

Thanks for your interest. We do generate our own workouts and currently don’t have an integration with Systm library, although we do have an integration for import/export with Wahoo.

Just give it a shot and see if you like it, the trial is free with no payment info required

2 Likes

I decided this year to use Xertonline Adaptive Training but good that there is competition.

I see you use the AlphaA1 metric to provide a readiness score during the warm up for a workout. two questions: How do you deal with this ii there are a lot of artefacts during the first 20 minutes which skew the data? And how is the workout changed that late is you have a poorish readiness to train?
Thanks

Good questions!

  1. We only do the readiness analysis if artifact corruption is less than 10% (for threshold determination we require less than 5% but here we can be a bit more lenient as it’s based on an average of all the data in the first 20 minutes, not just where the thresholds where determined).

  2. You have to make that call yourself if readiness is poor. We are doing this readiness feature in collaboration with alphaHRV (we have synced to use the same algorithm) but a) alphaHRV/your Garmin may or may not have access to internet and b) Garmin CIQ doesn’t allow a more sophisticated implementation at the moment (for example you could imagine having an easy workout synced and ready to go if readiness low)

Thank you. Do you have plans to alter the VT1 threshold at 0.75 in light of the recent research that appears to indicate that the threshold value can be very personal?

1 Like

I currently have a coach on Nolio, but I’m slowly considering a change.
I’m therefore interested in this forum.

Two questions:

  • I assume that AI Endurance calculates the different zones (power, heart rate, pace). Is this also exported to Intervals.icu?
  • My schedule is very variable due to my work as an emergency physician. Some weeks, I have four free days, and other weeks only two. Is it possible to input this into AI Endurance so the AI adapts?
    I imagine trying the free version would be easier, but for now, I’m injured, so I’m patiently waiting.

Thank you :slight_smile:

Thanks, subscribed!
too bad there’s no choice for run + cycling

3 Likes

We’re actually working on a bunch of stuff regarding the HRV thresholds right now. We’re actively involved in the research as well. I could go on forever about this so I’ll try to make it concise :slight_smile:

Right now where we’re likely going is

  1. less emphasis on ramps, for example Relationship of Cycling Power and Non-Linear Heart Rate Variability from Everyday Workout Data: Potential for Intensity Zone Estimation and Monitoring and https://www.medrxiv.org/content/10.1101/2023.12.19.23300234v1 indicate that simply collecting threshold data from everyday activities that are not necessarily ramp tests might paint a more representative picture of your thresholds/zones. We call these cluster thresholds.
  2. include respiration frequency (RF) into threshold analysis. https://www.medrxiv.org/content/10.1101/2024.08.14.24311967v1 achieved improvement on aerobic threshold estimation by combining a1 with RF (albeit in ramps)
  3. the potential overestimation/individualization of the 0.75 threshold: remains to be seen if that’s actually an issue once you go away from ramps. In ramps a big part of the overestimation might come from individualized cardiac delay (how long it takes your cardio system to catch up to the increased power/pace output). We often see 0.75 cluster thresholds are lower than 0.75 ramp thresholds.
  4. introduction of a baseline (power/pace where a1 > 1): can help to veto threshold detection. Baseline is also introduced in DDFA recently launched by Suunto, DDFA Dynamical Detrended Fluctuation Analysis | AI Endurance
3 Likes

We actually have a direct Nolio integration as well so that should work well. Regarding your questions:

  1. yes
  2. we do availability per week day (see screenshot) but you can also adapt your plan via drag and drop, skipping workouts or if you simply don’t do the workouts we might propose modifications depending on where you’re at in your plan/how crucial the workout was
2 Likes

Thank you for the subscription!

We actually do have run + cycling: choose ‘Triathlete’ user type and set your swimming hours/week to zero. Sorry it’s a bit hidden

1 Like

Hi Markus,

I wasn’t looking for a training plan but your app seems really interesting, so I decided to sign up.

One question: If I only have ~10 days of running power (just got Stryd pods) how will that affect the model?

Also a couple of notes while I waiting for my account to get set up. During the onboarding, I wasn’t given an option to select a goal race. The only options I got were critical running power or critical endurance power. I wasn’t able to add intervals.icu during the onboarding process either.

Regards,
Anthony

1 Like

Too much traffic? :thinking:

When I click on dashboard …


Apparently so. I have had this for a few hours. Apparently after the announcement here, many wanted to try it.

Hi Anthony,

In this case your model will be pre-trained with data from similar demographics (something you can switch on/off in the settings). Although in your case it’ll default back to pace until you have enough running power collected over a few weeks (basically because running power is messy ever since there isn’t just Stryd and Garmin etc have introduced ‘native’ running power the values are too different to reliably use a pre-trained model).

You’ll have more options to select races or combinations of disciplines to optimize once account is ready in Plan page and you can connect to Intervals.icu from Apps page https://aiendurance.com/dashboard/apps once account is ready.

Yes - sorry we are struggling a bit with the demand but have scaled up all our resources significantly so hopefully now all is working ok

Yes, sorry wait times are way longer than usual right now - thanks for your patience!

Hello!

The sync with Garmin stopped on 12/08 more than 3 hours ago…

There’s not data from Strava (Zwift rides), only runs from Garmin.-