Sönke Peters

Software Engineer @

[Sönke Peters @zunkp]

@conductor_build Says "Your annual license key to use Virtuoso Message List in non-production environments has expired. You can still use it in production. To keep using it in development, purchase a new key."

[Sönke Peters @zunkp]

@conductor_build your Virtuoso Message List license ran out 👀 Please let me work again.

[Sönke Peters @zunkp]

@billiesastre Yes same for me and others I’ve asked

[Sönke Peters @zunkp]

I'm so glad I diversified my personal AI agents so I could ask Grok to check if Instinct is actually down.

[telli @tellicom]

🚀 Day 5/5 of telli Launch Week

Introducing Browser Use: your voice agents can now use software like humans.

Show telli how to complete a task, and your agent can navigate pages, fill out forms, and get things done, even without a direct integration.

[Sönke Peters @zunkp]

Very helpful tool to find old agent sessions and have your agents ground themselves on findings from other sessions before they get to work https://t.co/Hny8RkZspq

[telli @tellicom]

🚀 Day 4/5 of telli Launch Week

Introducing improved Workflows: automate what happens before and after calls, all within telli.

And with the new Code Node, enterprise integrations are faster and more flexible.

So agents have the context they need, and customers get the follow-up they deserve.

[telli @tellicom]

🚀 Day 3/5 of telli Launch Week

Introducing Analytics: track your agents' performance and identify areas for improvement.

Start with a template, build your own dashboard, or ask Charlie to create one for you.

Then share your insights across the company with interactive reports.

[Sönke Peters @zunkp]

@benln telli (YC F24) is building the future of customer operations for B2C companies out of Berlin and hiring for engineering (Voice AI, Product), product design and more ☎️

https://t.co/Lu3Gne9ssy

[Sönke Peters @zunkp]

My agents probably hate me always spamming `/principle-never-block-on-the-human` when they ask stupid questions

[telli @tellicom]

🚀 Day 2/5 of telli Launch Week

Introducing Web Calls: test your voice agent directly in telli, without picking up the phone.

Talk to your agent and follow the transcript as it happens.

And when real customers call, you can listen in and follow those conversations live, too.

[telli @tellicom]

🚀 Day 1/5 of telli Launch Week

Introducing Simulations: automated testing for voice agents.

Hours of test calls, replaced by one prompt.

Tell Charlie what your agent must get right.

It creates the tests, runs them in parallel, and helps fix what fails.

[seb selassie @sebselassie]

our voice agents can use the browser now...

[Sönke Peters @zunkp]

Writing this from the office, waiting for the agent to let me go home

[Sönke Peters @zunkp]

I feel like working with agents has only made it harder to stop working. A few years ago pre-AI you'd get into a deeper flow, but you could just stop after finishing any step. Now you think "just one more prompt" and then wait another 45 mins for the agent to finally be done.

[Sönke Peters @zunkp]

Finally tried pstack this weekend. Been running five agents in parallel since I got to work this morning, all shipping PRs on their own.

The whole trust and verification thing works pretty well because the `/poteto-mode` main agent explains what it did and why that’s clearly the right solution.

[Sönke Peters @zunkp]

Waiting to try the OpenAI Decisions API be like

[Sönke Peters @zunkp]

@tellicom You can get the unredacted version here: https://t.co/qVL2bLNtkB

[Sönke Peters @zunkp]

When the stacked PRs have been sitting around waiting to be merged for a while @tellicom

[telli @tellicom]

Today we are launching telli Duo agents.

Over the past years, the industry standard for voice AI has been the cascading architecture.

Meaning when a human speaks to an agent, the voice gets transcribed by an STT (speech-to-text) model, goes into an LLM, and is transcribed back with a TTS (text-to-speech) model.

This worked well and kicked off the era of Voice agents. But it came with fundamental flaws: It's impossible for an agent to listen and speak at the same time.

A good conversation has a rhythm. We pause, interrupt, and respond before the other person has finished their sentence. It happens so naturally that we rarely think about it. Building this into an agent was impossible.

A couple of months ago we got early access to @OpenAI's GPT-live model. It made us rethink our entire Voice AI infrastructure and rework it to bet on full-duplex speech-to-speech models. The result was the new telli Duo agent.

We went deeper into how we did that in our blog post: https://t.co/hASQc0TV6b

PS: We know we over-exaggerate the difference in the video. The comparison is more with a traditional IVR bot. But it's still much better now, just harder to show in a demo than when you experience it yourself. Just give yourself a call on https://t.co/KsqwnodO7Y.