Charles Tenot

CEO @lemlist & lempire. I share what I learn as the CEO of a profitable SaaS growing to $40M to $100M+. Follow me to learn from the journey.

I love Tech, Sales, Marketing and Leadership. 

I share my learnings and thinking on LinkedIn, don't hesitate to follow & interact.

COO / General Manager

Growing lempire from 20M$ to 100M$ while generating 35%+ EBITDA

lempire products: lemlist, lemwarm, Taplio, Tweet Hunter, lemcal - more to come

I have no idea what my kids' life will look like with all this AI. 
But one thing I'm sure of: they need to understand that you can build things that other people use.

So I started vibe coding apps with them.

Last week, we built a cookie spot rating app to find the best spots in Paris.

This weekend, we made a Harry Potter sorting hat quiz (the "Sorting Hat" test).

Each one took about 30 minutes. 
And they're better than most apps you'd find online for the same thing.

I don't know what jobs will exist in 10 years. 
I don't know what skills will matter most.

But kids who grow up thinking "I can build this" will be in a better place than kids who just think "I can use this."

PS: sorry the test is in French

CEO @ lemlist & lempire

I just watched our PM build a complete feature in 2 hours.

No dev involved.

He used Claude Code to create a working mockup, test all the edge cases, then generated the PRD from the code.

This is how our entire Product team works now.

Here's the process:
→ They download the full codebase locally.
→ They use Claude Code with custom skills we built for coding best practices.
→ They vibe code their features.
→ Build the interface.
→ Test the flows.
→ Catch edge cases nobody thought of.
→ Then they use AI to generate the PRD from the working prototype.

What they deliver to Engineering:
→ A fully functional mockup that devs can test
→ A detailed PRD with user stories already mapped out

This reduces conception cycle by over 50%.

Better product decisions because they can interact with the actual feature.
Fewer bugs because edge cases are caught early.
Less confusion between Product and Tech.

Most Product teams are still writing specs and hoping devs understand.
We're shipping working prototypes before the first planning meeting.

This is what all Product teams will do in the future.

Curious to hear from other Product teams: how are you using AI in your process?

Help needed: if you are looking for a Senior Finance profile, my life partner, Violaine, is available now.

After spending 3 years building a network of kindergarten for children with neuro developmental disorders, she is now available for something new.

Why I recommend her, even though I'm clearly biased:

→ Worked with top Private Equity firms globally
Always recommended afterwards

→ Hybrid experience with Senior Finance roles (M&A, Reporting) and 3 years of entrepreneurship where she ran notably finance for their 50 people business.
She can be both strategic & hands-on, she's not the "ivory tower" CFO.

 → Amazing analytical skills, she notably worked on complex problems in manufacturing, much more complex than SaaS.

 → Great business sense, easy-going, high work ethics

What she's looking for:
→ CFO or Finance Director-level
→ Interim or part-time
→ Open to full time with flexible work location
→ Start date from March 9th

Don't hesitate to like or comment to help me get more reach, thanks a lot 🙏

→ Phone: +33 6 69 25 57 26
→ Email: vsaintromas@gmail.com

As a CEO, I'm almost never a decision maker, yet very few salespeople understand this.

I still get cold called, emailed, 10x per day.

But even when the approach is solid, the best I can do is forward it to someone on my team. 

And most of the time, I won't even do this. 
Just because I don't want to make my team think I have evaluated the solution and I recommend it, and ultimately I'm not the one who makes decisions for teams.

Teams make their own decisions.

The CEO role is about creating alignment, building the best team, and challenge them when needed. 

But I'm not the one who will say "we should buy this."

If your outbound strategy is targeting CEOs, it probably works up to a certain point. 
But past 100 employees, it might not be the best approach, unless you are actually selling to CEOs, but I can count on my hand the number of things that I decide to buy myself.

The person who feels the pain, evaluates the tools, and makes the call is almost never the CEO.  It's the C-Level, the VP or Head Of. 

My advice: don't waste your time on the wrong target.
Map your buying process, find the person who feels the pain, evaluates the options, and owns the budget.
Then spend your time preparing a great outreach for them.

undefined undefined

I read Cognism "Top 21 GTM Tools" blog post.

Clay wasn't in it. 
lemlist wasn't in it. 
Apollo wasn't in it. 
Salesloft wasn't in it. 
Outreach wasn't in it.
Gong wasn't in it.

The biggest names in GTM. 

All missing from a "top GTM tools" list.

That's not an accident.

When a company publishes a "best tools" list, they're not trying to help you. 

They're running a content play.

The rules are simple: 
→ Include tools that don't compete with you 
→ Include tools that make you look good by association 
→ Never mention the ones that could actually take your customers

So here's the irony.

If you're ON a competitor's "top tools" list, it means they don't see you as a threat.

If you're NOT on it, it means you're the one they're worried about.

Next time you read a "best tools" listicle, don't look at who's on it.
Look at who's missing. ✌️

Yesterday we ran a webinar about how ElevenLabs runs Go-To-Market at 330M$ ARR. 

Over 1,500 people registered. If you missed it and want the recording just drop a comment and I'll DM it.

Here's my human summary of what Jonathan Chemouny shared after 1 year there.

About ElevenLabs GTM:

→ 330M ARR, +300M in 2 years. Just valued at 11B.
→ Only 400 people. Close to 1M ARR per employee.
→ GTM team is about 100 people.
→ Org is SDR + AE but they call everyone "GTM." 
Limited titles, limited hierarchy.
→ They maximize distributed ownership per country and region. 
They hire founder-like profiles who run their region like CEOs.
→ Top performers come from Finance, even Product backgrounds (👋 Oscar Lanzendorf)
→ Their sales close an average of 20X their OTE 
(source: Carles Reina podcast at 20VC, link in comment)
→ They grew outbound from 5% to 46% of revenue in one year. 
Outbound lets them better target ideal customers and segments.
→ Average deal size ranges from 30K/year to millions for Enterprise.
→ Outbound sales cycle is slightly longer than inbound but average deal size is higher.
→ They receive thousands of inbound leads daily and built an AI agent that answers prospects 24/7.
→ Everyone does outbound, even if you're hired as a potential GM-level.

What enables their incredible success:
→ Very high hiring bar. 4 interviews, harsh panels.
→ They hire for smarts, not experience. 
They don't want people who come with playbooks. 
They want people who can build playbooks.
→ They use AI a lot but will never run things 100% AI. 
Humans think of a creative way to approach an account and craft the message. AI handles things like voice notes or AI-generated video.
→ They meet prospects and customers in-person a lot. 
Jon believes humans should do things AI can't, like building real relationships.
→ LinkedIn and WhatsApp are their primary outbound channels. Quality before quantity.
→ They don't monitor activity metrics. 
They track the number of meetings done with prospects.

And of course, they have an insanely strong product.

I hope you find this valuable. 

If you have a company or GTM person we should invite next, don't hesitate to share.

---

Bonus: we created an agent that was fed on Jon's knowledge and our webinar prep document, I'll share the link with the recording.

cc thanks Rémi Kokabi, Domitille de Saint-Exupéry Rafael Ornia Mitchell

You can beat 99% of companies by doing a few things well:

→ Set one clear vision and don't change it every 3 months 
→ Say "no" to everything that doesn't contribute to the strategy 
→ Remove low performers after weeks, not years 
→ Use your own product 
→ Talk to customers 
→ Keep reporting light 
→ Lead every meeting from the customer's standpoint

It looks simple written down like this.

But most companies don't have an execution problem. 
They just keep adding layers, processes, and distractions on top of it.

The companies that win aren't doing anything secret. 
They're just not getting in their own way.

I asked someone at ElevenLabs what makes their culture different. 
Their answer helped me understand how they went from 30M to 330M ARR in less than 2 years.

Yanek Brinkmann, one of their GTM team, told me:

"It's the closest you can get to doing your own business while still being in a job."

I've always believed that someone who works like they're running their own business has 10X more impact than someone who's just doing their job.

The difference is ownership. Real accountability. 

The kind you only get from people who actually care.

You'll get more from someone less experienced with insane grit than from someone senior who's just collecting a paycheck.

ElevenLabs understood this. 

And they still operate like that at 400 people.

Imagine 100 GTM people with a founder mindset.
→ Clear ownership on their region. 
→ Working directly with Product to adapt when needed. 
→ Flat org. No titles.

Their reps close 20X their OTE. The industry average struggles to hit 5X.

But this only works if you actually distribute ownership and responsibility. 
Not just talk about it.

I'm convinced this is how world-class companies are built now. 
And they will crush big groups with pyramidal organizations.

➡️ Later today, we're sharing more about ElevenLabs's playbook with Jonathan Chemouny. 

Already 1,350 people registered. You can still take your spot, it's free value.
No recording. We want people live and contributing.

Link in comment.

Really happy to see lemlist ranked #1 software in France on G2.

Not because of the ranking itself but because of what it took to get here.

The last few months have been the strongest innovation pace we've ever had. 

And all of it comes from listening to our 20,000+ customers every single day.

Sales is still one of the hardest jobs in most companies. 
No product has made it easy yet. Including ours.

But that's exactly why we keep building.
Let's keep going.

Big up to the team ❤️

I have data-backed proof that spray & pray doesn't work.

It comes from 350M+ emails and 4.5M LinkedIn messages sent with lemlist over the past months.

This chart represents our 20,000+ customers.

From pre-revenue founders to companies with 300M$+ in ARR and 100+ sales reps.

How to read it:
→ Horizontal axis: our 10 deciles of customers, from the 10% with the worst results to the 10% with the best
→ Each bar: the volume of messages sent
→ Vertical axis (line): the average reply rate

What we see:
There's a clear negative correlation between volume and reply rate.
The fewer messages teams send, the higher their reply rate.

In detail:
→ The first group shows that the bottom 10% aren't sending enough messages, or just lack adoption. They'll probably churn unless they use lemcoach.
→ Then across the other 9 deciles, each one has lower volume and a higher reply rate than the previous one.

If this doesn't show you that it's better to do less but better, I don't know what will.

Spoiler: it's not the spray & pray "AI SDR Agent" that will solve this.

Every Tech company board is the same: "We need to go Upmarket"

The reason is usually financial. 

Going upmarket increases "Enterprise Value", because is usually means higher retention.
And of course, investors want to increase enterprise value.

But going upmarket at the expense of losing your historical market isn't always the right call.

You might lower churn. 

But you can lose growth and end up struggling.

This is why at lemlist, we spent over 9 months debating if we should go upmarket or not. 

Because serving small customers was always important to us for many reasons:
→ They trusted us first
→ More customers = bigger brand
→ Small companies can become big ones with time
→ Serving emerging businesses drives real purpose 
→ Startups are usually the innovators, they push us to keep improving the product

We finally decided to build lemlist for sales teams and go upmarket, but we decided to do it slowly, not as a pivot, but as a continuous progress of the platform. 

And it paid out because today we have customers from 1 to 150 users.
We have grown by +70% in 2025 while being profitable, and across all segments.

After seeing from the inside 4 different companies go upmarket, I wanted to share my learnings:
Wednesday with Lucas Bédout from Hyperline, we are organising a meet-up in Paris.

Here's a preview of what we'll talk about:
→ Why it took us 9 months to decide between going upmarket or continuing to build a suite for founders 
→ The stages we went through, mistakes we made, and what we did right 
→ Key people we hired (hey Yann Guilleux) 
→ How we run inbound & outbound, and how we use our own product 
→ How we grew the sales team from 1 person (hey Tal Baker-Phillips) to 30+, across 3 areas and 2 roles (AE & Inside Sales) 
→ How we signed our first 100K+customers, coming from 1K$ARPA 3 years befre
→ How we monitor productivity & measure performance 
→ What we built in the product to maximize conversion & score leads 
→ How we organized RevOps & Tools 
→ Our Sales Quota, Payback and how we maintain <3 months payback while adding Sales headcount

Bonus: how we increased ARPA by +30% in one month by changing the wording of one question in our product onboarding 🔥 

Free event. Paris. March 11.
Link in comments.

cc: thanks Antoine Leprince for the organization 🙏

Slack, Figma, Notion, Dropbox, Datadog. They all started PLG. 
None of them stayed PLG. Here's why:

PLG is great for small customers. 

Low CAC, high scalability, self-serve magic. 

But as your product improves, you start attracting bigger companies. 
And enterprise buyers don't convert on PLG products. 
They need help. They need sales.

This is where PLG + Sales-Assisted comes in. 

And I'm convinced it's the ultimate state for SaaS:
→ PLG for small customers. Low CAC, fully self-serve, scales beautifully.
→ Sales-assisted for larger customers. You can segment by geo, industry, company size and organize your team around it.

This combination maximizes conversion across every customer segment.

Since we added a sales motion at lemlist, our growth has never been stronger. 

We grew +70% last year. PLG + Sales-Assisted together.

But here's the hard part:

In PLG, everything is streamlined: credit card payment, automated provisioning and payment retries.

But when sales gets involved, you start dealing with:
→ Custom quotes
→ Custom terms and legal discussions
→ Discounts and approval workflows
→ Custom billing terms and payment methods

You go from a scalable model to a non-scalable model overnight. 

And that's where most companies struggle.

The good news is that with the right tech, you can now build scalable systems even for sales-assisted motions.

On March 11, I'm co-hosting an event in Paris with Lucas Bédout (CEO at Hyperline) and the Business Operations Network to dig into exactly this. 

How to build a hybrid revenue engine that scales.

Free for BON members. 

Link to register in comments.

cc Victoria Dalleau 🙏

Nothing prevents you from sending 10 follow-ups. 
And yes, you'll get more replies.  But it's doing invisible harm to your outreach.

If you look at reply rates by number of follow-ups, the pattern is clear. 
Each follow-up increases your chance of getting a reply. Even a little.

But the incremental gain drops fast. 
After 4 follow-ups, you're barely moving the needle.

And here's what most people miss.

Every extra follow-up increases the chance your leads mark you as spam. 

Not by replying. 
By hitting that little spam button because they're annoyed.

When that happens, it hits your domain and email reputation. 
Your emails start landing in spam more often. 
Not just for that lead. For everyone.

This is the classic trap where short-term gains quietly destroy long-term results.

More follow-ups today = Fewer emails delivered tomorrow.

As a CEO, when people agree with me too easily, I don't take it as a compliment. I take it as something to question.

Things I don't expect from people in the team:

→ Telling I'm right
→ Liking my LinkedIn posts
→ Being overly positive on my ideas and initiatives
→ Nodding in meetings
→ Sugarcoating bad news
→ Using my notes / inputs as is without re-working them
... and certainly not telling me I'm smart, or giving any compliment

Here's what I actually value:
→ Saying what you think
→ Thinking on your own before defaulting to the group
→ Coming to me when you disagree or don't understand, especially when it's uncomfortable
→ Bringing problems early and as they are
→ Having a point of view and defending it

I recently listened to a podcast from an experienced CEO who said his biggest mistake was surrounding himself with people who gave him compliments but weren't truly competent at their jobs.

The best people I've worked with were never those who constantly agreed with me, they were the one who made me change my mind.

We’re organizing a webinar with of the brains behind Ramp’s GTM systems, now building Perplexity’s Enterprise growth engine. 

Nathan spent 3.5 years at Ramp building the business systems behind one of the most successful sales-led companies in the world.

Now he's Head of Enterprise Ops at Perplexity, building the GTM machine of an AI company. From the inside.

I'm really excited about this one. Nate is an absolute GTM master, and we're very lucky to have him.

There are hundreds of LinkedIn posts telling you to comment "Claude" to get an AI config that runs your GTM on autopilot. 
We've tried many of them and the results are always disappointing.

Over the past months, I've interviewed 50+ companies to understand how they actually use AI for sales. 
The reality is far from what you see on LinkedIn, most sales teams are still doing things manually, with a ChatGPT tab open on the side.

Nate is different: he actually built AI-powered GTM systems that people use every day, and that deliver real ROI. 

Here's a preview of what he shared with us:
→ Reps used to spend 30 min prepping each account. 
Now they trigger deep research in natural language and get a full brief in minutes.

→ Outreach used to be {first_name} + generic template. 
Now it references real strategic context. 
Earnings calls, hiring signals, tech stack changes.

→ RevOps used to own every playbook. 
Now the rep is the operator. No bottlenecks. No stale processes.

→ And some things they tried but decided not to release, usually because AI can prep you, but AI can't be you 👀 

He's going live with us to walk through the exact workflows. 
Not theory, but how things actually runs in production.

If you lead Sales, RevOps, or GTM, this one's worth your time.

Link in comments.

VIP hosts: Pierre Touzeau and Rémi Kokabi  🔥

We have 2 data analysts at lemlist, but one is paid in tokens and never sleeps. 
It's's called Squilly. 

Squilly is our AI data analyst. 

It runs queries, builds dashboards, and answers data questions from anyone on the team. 24/7.

You don't have to wait that someone has time to handle your query.

What's good with Squilly is that he qualifies your need before answering.
As you can see in the video, it always asks what exactly I'm looking to know.

You can see that it shows you the query & the cost too, pretty useful.

Squilly now handles a significant share of our data requests.
It's pretty cool because it manages simple, low value requests and our Data team can actually focus on more complex tasks. 

Employees have answer faster as it removes the ticketing logic.

It's the same pattern I already shared with AI managing 56% of our support tickets, AI does a great job to manage most simple tasks, leaving human with more value-adding things to do. 

Big up to the Data team 👏

This week, a bot fixed its first production bug at lemlist. 

Here's what happened:

A user double-clicked a delete button. That sent two requests at the same time.  The API returned a 409 conflict error.

Our bot detected it in the logs, investigated the root cause, traced the code path, wrote the fix, and submitted a pull request with the right reviewers assigned.

Total human involvement: reviewing and merging the PR.

In the meantime, we also had 3 features shipped to production by... Product Managers.

A year ago, we never thought this would happen so fast.

I was talking to a friend, a developer with 20+ years of experience.
He told me his workflow has completely changed.

Before, he spent most of his time in "flow." Going from one coding step to the next.
Now, he spends more time thinking about the overall architecture. 
Then he works with AI to develop the different parts. Then he reviews the code. Makes sure the logic is right.


And yet, we're still hiring developers.

Because our bet is that the job isn't disappearing. It's shifting. 
And good developers will always be a scarce resource required to build great products.

You could argue AI will improve so fast it won't need developers at all.

My gut says you'll always need people who understand systems deeply enough to work with AI.

But we'll see. What's your take?

Sending a LinkedIn message in a sequence increases email reply rate by 38%. 
And no, that's not a typo. I'm talking about email replies. 

LinkedIn reply rates are averaging 15%, vs. 1-3% for emails (up to 15X higher)

But here's what we found across 35M leads:

When you add LinkedIn steps to an email sequence, your emails start getting more replies too.

The prospect sees your name on LinkedIn. 
Maybe they check your profile.
Maybe they read your message.

Then your email lands, and you're not a stranger anymore.

That familiarity is worth +38% on your email reply rate.
This data is coming from 35M leads that we analyzed at lemlist.

This is pure facts that multi-channel is the right way to do outbound today, and why we build lemlist to include as many channels as possible.

---

6 examples why short-term growth almost always comes at the expense of long-term growth:

→ Hiring the least worst candidate because you need someone now vs. waiting for the best even if it takes 6 more months. 
Impact compounds. An A-player generates 100X more value than a B-player.

→ Sending 10 follow-ups to squeeze out more replies vs. stopping at 4. 
Each extra follow-up barely moves the needle but increases spam reports. 
Your domain reputation takes the hit and deliverability drops for everyone.

→ Blasting your entire TAM in one month to hit pipeline targets vs. working accounts in waves. 
You get a short spike, then you've burned your brand with every account.

→ Cutting product investment to boost short-term margins or hire more Sales. Customers don't leave immediately. 
They leave 12 months later when a competitor ships what you didn't.

→ Discounting to close a deal before quarter end vs. holding your price. 
You hit this quarter's number, but you've trained the buyer to expect discounts. Next renewal, they push harder. 
They talk to their friends. 
And your average deal size erodes across the whole pipeline.

→ Not promoting a great rep with manager potential because they're bringing in short-term revenue. 
They'll leave 6 months later. And you'll end up hiring someone less good.

Every one of these looks like the right call in the moment. The cost just shows up later.

That said, there are moments where you need to move fast and accept some debt. 

The interesting part is knowing when you're making a conscious trade-off vs. when you're just not seeing the long-term cost.

Most outbound teams find out about deliverability issues weeks too late. 

By then, the damage is done, your pipeline is already melting.

Then what I call the "circle of short pipeline" starts:
→ Reps have fewer deals
→ They desperately want to close the ones that are open
→ They start looking desperate
→ Prospects feel it
→ They ghost you
→ You lose deals you could have won
→ CRM reason: "no decision"

This is why fixing deliverability issues fast is critical.

But the problem is getting visibility on what's actually happening. 

Deliverability depends on multiple factors: technical setup, domain and email reputation, copy, data quality, bounce rates, and more.

This is why we just shipped the Deliverability Hub in lemlist.

One place to see everything:
→ Delivered vs. spam for warm-up emails
→ Delivered vs. bounced for outreach emails
→ Deliverability scores by mailbox and domain
→ Reputation issues spotted early, before they tank your results
→ Filters by timeframe, mailbox, and provider

No more guessing. No more juggling. 
Everything is in one place.

We already have great feedbacks about it, kudos to the team 🔥

CTO at lemlist - building the best place for devs

👨‍💻 CTO with 20+ years in tech—from C++ in Taiwan to AI agents in France.

CTO focused on building high-performing teams and delivering real business impact through tech.

After starting my career coding in C++ and .NET in Taipei, I spent over a decade managing large-scale projects at Airbus, before diving deep into the startup world.

Today, I lead engineering at lempire, where we build sales tools like lemlist, lemwarm, and lemcal. I joined as a fullstack dev passionate about Meteor.js—and grew into the CTO role by focusing on scale, simplicity, and impact.

I love:
✅ Lean dev processes that actually deliver value
✅ Empowering devs to think like entrepreneurs
✅ Keeping tech aligned with real business goals
✅ Still getting my hands dirty in code when it matters

Whether it’s building teams, launching products, or simplifying complexity, I bring both the mindset of a developer and the vision of a product leader.

Mika F

Claude Code va tuer votre lock-in

Je suis diabétique de type 1. Je porte un capteur de glucose FreeStyle Libre 2 d'Abbott sur le bras 24h/24.

Je voulais juste lire ma glycémie depuis mon Mac. Un truc simple. Mes données, mon capteur, mon corps.

J'ai passé des semaines avec Claude Code à reverse-engineerer le protocole BLE d'Abbott. 15 tentatives. Fake JNI, crypto propriétaire, bibliothèques ARM chiffrées. Résultat : Abbott a tellement verrouillé son protocole que même en y mettant des moyens sérieux, c'est un mur.

Alors je vais faire le truc le plus simple du monde : je vais demander à mon médecin de me prescrire un Dexcom.

Pas parce que Dexcom est techniquement supérieur. Mais parce que leur protocole BLE est documenté par la communauté open-source, que xDrip+ le supporte, et qu'en 2 heures avec Claude Code je peux me construire l'app que je veux.

Et c'est là que ça devient intéressant pour les industriels.

Avant, le lock-in marchait. Reverse-engineerer un protocole BLE demandait des mois, des compétences rares, une motivation hors norme. Le coût de sortie était prohibitif.

Aujourd'hui, n'importe qui avec Claude Code peut tenter de construire son alternative. Et quand il se heurte à un mur propriétaire, il ne s'acharne plus. Il change de fournisseur. Le coût de sortie est tombé à zéro.

Le nouveau calcul est simple :
- Protocole ouvert + API → les devs construisent votre écosystème gratuitement
- Protocole fermé → vos utilisateurs les plus engagés partent chez le concurrent qui ouvre

Abbott, Medtronic, et tous les autres : ouvrez vos protocoles. Proposez des API. Laissez les gens accéder à leurs propres données de santé.

Sinon, Claude Code fera le tri pour vous. Et ce ne sera pas en votre faveur.

 #WeAreNotWaiting #DiabetesType1 #ClaudeCode #MedTech #OpenHealth

Un ticket Notion.
3 agents IA.
Une PR créée sans humain dans la boucle.

Et non, ce n’est pas un hackathon.
On vient de mettre en prod un système multi-agents qui prend un incident réel, l’investigue (logs + code), se review lui-même, fixe, crée la PR et documente le ticket.

La partie intéressante n’est pas le “bug fix automatique”.
La partie intéressante, c’est l’orchestration :
– rôles stricts
– permissions limitées
– boucles de validation
– mémoire partagée contrôlée
– arrêt automatique si le système doute

C’est de l’ingénierie d’agents.

Post viral d'Eric Holmes : "MCP is dead. Long live the CLI."

Son argument : les LLMs n'ont pas besoin d'un protocole spécial. Donnez-leur curl et ils se débrouillent.

Il a raison.
Et ça tombe bien : lemlist est déjà en CLI.

curl https://lnkd.in/eC7GiwYR \
  -H "Authorization: Bearer lem_xxx" | jq '.[].name'

Lister vos campagnes, ajouter des leads, lire vos stats.
Zéro SDK. Zéro MCP. Juste curl.

Votre agent dans Claude Code ou un script bash peut piloter lemlist aujourd'hui.

Et pour ceux qui n'ont pas de terminal (Claude Desktop, Cursor, des sales qui veulent un chat) on a aussi un serveur MCP.

Le débat MCP vs CLI est un faux débat.
La vraie question : est-ce que ton produit est scriptable ?

API REST = CLI-ready.
MCP = chat-ready.
Les deux = agent-ready.

lemlist a les deux.

Coder est devenu rapide, presque trivial avec l’IA.
Le vrai coût n’est plus d’écrire du code, mais de choisir quoi construire.
Quand tout devient possible, la ressource rare n’est plus la technique, c’est le focus.
Démarrer un projet prend quelques heures, le terminer demande toujours des mois.
La différence ne se fera plus sur la vitesse d’exécution, mais sur la qualité des décisions.

Dans ce contexte, savoir “tuer vite” devient une compétence stratégique. Il faut accepter d’arrêter un projet même s’il fonctionne techniquement, s’il n’apporte pas assez de valeur. Prioriser dur signifie dire non plus souvent que oui, limiter volontairement le nombre de chantiers, et protéger l’attention de l’équipe. Sans cette discipline, l’IA n’accélère pas le progrès, elle accélère la dispersion.

En 2026, mon workflow quotidien :

Terminal custom (Kova) Rust + Metal
Timer (pomo) CLI Rust
Radio (cliamp) CLI
Notes & tâches (Panorama) CLI + MCP
Code (Claude Code) CLI

Pas d’Electron.
Pas de dashboards.
Pas d’apps de 200 MB qui font une seule chose.

Même mon timer pomodoro est en CLI.
Même ma radio est en CLI.

Pourquoi ?

Parce que je veux un environnement homogène.

Une seule surface d’interaction.
Les mêmes primitives partout : commandes, pipes, scripts.

Quand tout passe par le terminal, un agent IA peut faire exactement ce que je fais.
Pas besoin d’API spéciale.
Pas besoin d’intégration dédiée.
Il utilise les mêmes outils que moi.

Andrew Karpathy parle de “legacy technologies” qui deviennent “agent-native interfaces”.

Le terminal Unix (années 70) était pensé pour la composabilité.
En 2026, c’est précisément ce qui le rend compatible avec les agents.

La techno la plus ancienne devient l’interface la plus universelle.

Moins de couches.
Moins de friction.
Plus de scriptabilité.

Peut-être que le futur du software n’est pas plus visuel.
Peut-être qu’il est plus uniforme.

Si tu veux maximiser ton impact, pour toi et pour tes agents, recode tes outils en CLI, ou au minimum rends-les scriptables et composables.

taplio

The Battle

Charles Tenot

Mika F

Followers

Posts

Engagements

Likes

Performance ScoreCharles Tenot's LinkedIn Performance

Performance ScoreMika F's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Charles Tenot

Mika F

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Charles Tenot's LinkedIn Performance

Performance ScoreMika F's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreCharles Tenot's LinkedIn Performance