Cykel AI

Building the world's most capable digital workers.

[Cykel AI @CykelAI]

👀

[Sam Altman @sama]

10 am livestream today to launch a new product I'm quite excited about!

[Cykel AI @CykelAI]

5/5 Key Takeaways

✅ Intent signals are the biggest differentiator
✅ Deliverability is table stakes
✅ Multi-channel wins
✅ Automation ≠ intelligence

Want the full “AI Sales 2025” report? Drop a comment 👇

[Cykel AI @CykelAI]

AI sales tools exploded last year, but the market’s fragmented.
We analyzed 50+ companies across automation, outreach, and intent data.
Here’s what you need to know 👇

[Cykel AI @CykelAI]

4/5 Buying Signals + Channels

Better intent data = better timing & personalization.

Top platforms combine:
Email + LinkedIn + CRM data
⏱️ 24/7 signal monitoring

62.5% now support multichannel outreach

[Cykel AI @CykelAI]

“Oh you know, things are good, just wearing eight hats.”
Every founder, until 2025 when agents become part of the workflow.

[Cykel AI @CykelAI]

https://t.co/DzOdXHcgfD

[Cykel AI @CykelAI]

Introducing GTM AI: Zero-setup sales automation that automatically understands your business and creates sophisticated outreach campaigns instantly. No technical configuration required. #GTM #SalesAutomation #CykelAI
↓

[Cykel AI @CykelAI]

High application volume can overwhelm hiring teams. See how @PaidiaGaming could have addressed this challenge and improved screening efficiency with Lucy AI.
Case study insights: https://t.co/uSGwYn1DtJ

#AIrecruitment #HRtech

[Léo Mercier @leomercier]

Sentiment detected Tao last week (again), and this time looking into it with great depth.

Bittensor is complicated - and still grasping all sides.

But I have put below how I am thinking about it is so far (it may change the more I read and as I look through more code)

1. It’s a simple core blockchain itself - that gets validated and blocks produced. This only provides consensus and rewards.

2. The rewards aren’t to run the blockchain - but incentives the utilization of it. The blockchain emits rewards via Tao into areas where it is used. These use cases are called subnets. Currently only technical subnets exist (currently about 100) AI infrastructure, storage, image classification, and then some data analytics.

3. Each subnet is its own world - and anything can happen in it. The blockchain doesn’t care about what happens on a subnet and data isn’t computed onto it. There is no impact of scaling due to activity in a subnet.

4. There are validators and miners within each subnet - they perform the work and then agree what was correct. They use mostly use AI to agree on quality. For example was the LLM response correct or not for the users query?

The miners and validators test each other by also working on AI to AI generated sample user content feed with real user content so the two are indistinguishable. This allows them to work and fine tune themselves faster.

5. Only the top 64 validators earn reward tokens in a subnet. This forces a race of best results in each subnet. Similarly to the subnets themselves the more stake of Tao from delegators (users) the more rewards they’ll receive.

6. Unlike Solana you can gain and lose by staking. You stake your Tao into a subnet- and that converts your Tao into what is called an alpha token. The exchange between the two is handled by a AMM (automated market maker) contract with a two sided ratio of Tao in the pool to alpha token left. Same as how all the DEXs currently work. The more people stake into the subnet the more expensive the conversion is from Tao to alpha. You have a first in advantage same as other tokens.

7. By staking you earn a proportional distribution of 41% of emissions (paid in alpha - the subnet token) and rewards (paid in Tao to the pool) of the subnet.

8. Creating a subnet yesterday would have cost 355 Tao so around $100k - only one can be created per day. The direct reward for a subnet is the 18% distribution of the rewards and emissions.

9. Owning subnet is likely to be by teams that will use the underlying output of the work of the subnet.

For example if you had an end user application like a FaceSwap tool.

Reasons the team would use a Bittensor subnet.

- don’t want to have one large model providing all the results of the face swapped images
- don’t want to manage the infrastructure and scaling issues related.
- crowdsource updates to the models and improvements in quality to the FaceSwap output.
- focus on the user application only and allow others to be incentivized for the models.

The LLM models and inference would be done via the subnet in defi-ai manor (ai generated output - text, video and audio).

The miners (LLM or GPU providers) would run the models and compete for the most cost efficient way to provide FaceSwap the output. Remember only top 64 get rewards.

There is a lot more going on and will continue reading. One thing I have found is there is an EVM layer to it - which allows transfers, smart contracts but haven’t seen anyone talk about this yet.

Comment questions or corrections - and I’ll write another summary.

[Cykel AI @CykelAI]

"The perfect candidate doesn't exi—"
Every recruiter has said it. The unicorn candidate just doesn't exist...
Until Lucy finds them hiding in your ATS.
Keyword searches miss great candidates. Lucy analyzes beyond keywords to find your hidden gems.
https://t.co/U4HymYuXGL
#AIrecruiting #HiringWithAI

[Ewan Collinge @EwanCollinge]

I'm excited to announce that @CykelAI will be attending the @recexpo in London this Tuesday and Wednesday.

If you're attending and want to connect with us, send me a DM!

[Ewan Collinge @EwanCollinge]

"We lost a brilliant engineer because they sat in our ATS for a week without anyone noticing. By the time we reached out, they'd accepted elsewhere."

We hear stories like this a lot.

The reality is that high-quality candidates are typically off the market within just 10 days.

[Ewan Collinge @EwanCollinge]

How do you teach an AI to screen CVs?

Since people have been asking, here's a breakdown of how Lucy, our AI recruiter, screens CVs like a human.

It's more intuitive than you might think. 🧵

[Cykel AI @CykelAI]

Your recruiters: 23 hours screening CVs per hire.
Lucy: 23 seconds.
While Lucy screens candidates 24/7, your team can focus on connecting with top talent.
Greenhouse, Ashby and Teamtailor users try Lucy FREE: https://t.co/c7ZCJRWyVm
#AIrecruiting

[Cykel AI @CykelAI]

Every recruiter's question about AI.
Lucy isn't replacing recruiters—she's elevating them. She handles CV screening so your team can build relationships with top talent.
Using Greenhouse, Ashby or Teamtailor? Try Lucy FREE and cut screening time by 90%: https://t.co/c7ZCJRX6KU

[Ewan Collinge @EwanCollinge]

Lucy now works with @Greenhouse!

Here are some of the useful things she can do with access to your Greenhouse: 🧵

[Ewan Collinge @EwanCollinge]

Excited to share the first glimpse of Samson, an AI research agent we've built at @CykelAI

Give Samson a research topic and he'll search the public internet and private data sources to perform in-depth, high-quality research – similar to a junior analyst in a consultancy.

[Ewan Collinge @EwanCollinge]

Our AI recruiter can now accurately predict who you'll hire!