Christ First | AE @Airtable | Host - MSP Revenue Secrets

Rhondre Giscombe

Disciple of Christ | GTM @Airtable | Building apps & agents for Operators

VP, Head of Product Management @ Airtable

Airtable's MCP server is now live, along with connectors for both Claude and ChatGPT. 

Airtable is increasingly becoming a place where both humans and agents collaborate. By connecting agents to Airtable, you can allow them search and analyze your data, create records and update existing records. 

PMs now frequently ask their agents to synthesize product insights, customer feedback, and roadmap plans stored in Airtable apps -- or even feed this data directly into coding agents to build new features & prototypes! 

We'll be investing heavily in additional MCP capabilities over the coming weeks. Stay tuned for more updates including: 
- Updating Airtable apps via MCP (create new tables and fields)
- Support for users with Interface only access 
- Improved throughput and rate limits 

What additional capabilities would you like to see? Let me know in the comments! 

Try it out: https://lnkd.in/gZiCJtSm

WE HELP businesses across various industries automate their workflows and integrate their applications seamlessly, enhancing productivity and efficiency.

How WE DO that:
- Provide a no-code platform that allows users to create custom workflows without programming knowledge.
- Support full-code options for advanced customization and flexibility.
- Offer ready-made integrations for popular apps and services, from Microsoft 365 and Google Workspace to Slack and Discord.
- Include a wide range of AI models ready for instant use, allowing businesses to easily leverage artificial intelligence.
- Enable users to create conditional logic and branching scenarios for complex workflows.
- Ensure smooth and fast onboarding with our expert support team.
- Provide continuous updates and additions of new apps to our integration library.

Working with us, YOU WILL:
- Achieve higher efficiency by automating repetitive tasks and streamlining operations.
- Save time as our platform simplifies the integration and automation process.
- Enhance productivity with our user-friendly interface and robust feature set.
- Gain access to a wide range of pre-built integrations and the flexibility to add custom ones.
- Benefit from our dedicated support and extensive resource library to get the most out of our platform.

Oleg Zankov

Programming in 2026 is the new math. Let me explain.

We spent years learning math. But in real work? Nobody calculates in their head. Everyone uses Excel and calculators.

Yet give the same spreadsheet to someone who understands math vs. someone who doesn't. Same tools. Completely different outcomes.

Math was never about computation. It was about thinking.

Same thing is happening with programming right now.

Yes, anyone can build an app with Cursor or Claude Code. But I've watched two people tackle the same task with AI agents — one knew programming, one didn't.

One got a production-ready system. The other got a demo that explodes when a real user touches it.

The programmer wasn't writing code. They were thinking in systems. The non-programmer was just vibing.

And the flip side: if you're a developer refusing AI agents because "I can write it myself" — that's ego cosplaying as principle. You're doing long division on paper because you don't trust calculators.

Programming didn't die. It evolved from a trade skill into a thinking skill.

Just like math.

Which side are you on?

Founder @ Latenode.com & Speaknode.com | Studying how teams become 10x more productive with AI | Workflow automation & Voice AI

IBM just lost 13% of its market cap in a single day. Worst stock crash since 2000.

The trigger? Anthropic announced Claude Code can now map COBOL dependencies, document business logic, and identify modernization risks automatically.

Work that took human architects 6-12 months now takes hours.

Think about what that means.

For decades, enterprises paid billions to maintain systems everyone claimed were "too complex to touch." Too mission-critical. Too deeply embedded.

But COBOL powers 95% of ATM transactions. Airlines. Financial cores. The systems aren't complex - they're just undocumented.

The real moat was knowledge hoarding.

Consultants who memorized spaghetti code. Vendors who kept documentation intentionally sparse. An entire $200B ecosystem built on one thing: making sure you couldn't leave.

AI just made that friction disappear.

The companies now scrambling aren't the ones with complex systems. They're the ones who built empires on your inability to understand your own infrastructure.

If your vendor is telling you modernization is "too risky" - ask yourself who that risk actually protects.

#LegacyModernization #EnterpriseIT #AIDisruption #COBOL #DigitalTransformation

The AI with the highest benchmark scores is losing the enterprise market.

The smartest AI model is losing.

Gemini 3.1 Pro just doubled its reasoning benchmark score. 77.1% on ARC-AGI-2. Destroys Claude. Destroys GPT.

And it doesn't matter.

Here's the part Google won't put in the press release:

Anthropic's models still dominate GDPval-AA — the benchmark measuring real-world expert tasks in finance and legal work.

The tasks enterprises actually pay for.

I've watched this pattern repeat across dozens of AI procurement decisions. The team obsesses over leaderboard rankings. They switch to the "smartest" model. Then output quality drops on the work that matters.

Why? Benchmarks measure abstract reasoning. Enterprises pay for reliable execution on specific high-value tasks.

These are not the same skill.

A model can score 77% on reasoning puzzles while being worse at the $500/hour contract review that justifies your AI budget.

Benchmark leadership and market dominance have officially decoupled.

This means your "which model is best" conversation is broken. You're optimizing for the wrong scoreboard.

The question isn't "which model is smartest."

It's "which model is smartest at our actual work."

Who else has seen the highest-scoring model underperform on real tasks?

Everyone's chasing smarter models. But for those of us building AI products — speed is the real bottleneck.

Inception Labs just dropped Mercury 2 and it changes the math completely.
1,009 tokens/second. Not a typo.

How? They ditched autoregressive decoding entirely. Mercury 2 uses diffusion — generating multiple tokens in parallel instead of one at a time.

Why I'm excited as someone building AI agents and automation:
→ Voice AI: reasoning-quality responses within real-time speech latency 
→ Agentic loops: 5x more steps in the same time budget 
→ RAG pipelines: add reasoning without blowing latency 
→ $0.25/1M input tokens — cheaper than most alternatives 
→ OpenAI API compatible — drop-in replacement

This isn't a better model. It's a different architecture.

And it unlocks product experiences that were simply impossible before.

The autoregressive era gave us intelligence. The diffusion era gives us intelligence at production speed.

Go try it → chat[dot]inceptionlabs[dot]ai

What would you build if latency was no longer the constraint?

"Vibe coding" is killing AI adoption in engineering teams.

Not the practice. The term.

Here's what happens:
You walk into a standup with senior engineers and say "we should try vibe coding."

The room goes cold.
Eyes roll.
Someone smirks.

Because "vibe coding" sounds like something a YouTuber does in a hoodie at 2 AM while whispering "let's see if this works."

It does NOT sound like something a principal engineer at a fintech company should adopt to ship faster.

And yet — the methodology behind it is genuinely transformative:
→ AI-assisted architecture decisions
→ Automated boilerplate generation
→ Intelligent code review and refactoring
→ Rapid prototyping with human oversight

This is serious engineering leverage. But we wrapped it in a joke name.

Imagine if "pair programming" had been called "buddy coding." Would any enterprise have adopted it?

We need a professional term. One that engineering leaders aren't embarrassed to put in a RFC or a quarterly roadmap.

Here are my three candidates:

1. Augmented Engineering — you're still the engineer. AI augments your capabilities. Clear, professional, and accurate.

2. AI-Paired Development — borrows credibility from "pair programming," a practice every senior dev respects. Your pair just happens to be an LLM.

3. Agentic Development — aligns with the broader "agentic AI" movement. Signals that AI isn't just autocomplete — it's an active participant in the dev workflow.

The winner matters more than you think.

Terminology shapes perception.
Perception shapes adoption.
Adoption shapes who gets left behind.

The teams that figure out the RIGHT framing will move 2-3x faster. The teams that dismiss this as "vibe coding for juniors" will wonder why they're losing talent and velocity.

---

Which term would YOU use to pitch AI-assisted development to a room full of skeptical senior engineers?

Drop your term in the comments. Bonus points if it's something your CTO wouldn't cringe at. 👇

undefined undefined

Co-Founder & Product Owner at Latenode.com & Debexpert.com. Revolutionizing automation with low-code and AI

Clay’s pricing change is a good reminder for SaaS founders.

When you build your internal systems on a single vendor, their roadmap eventually becomes your roadmap.

Clay didn’t just change pricing.

They changed the economics of running GTM workflows at scale.
And that’s completely normal.

Every successful SaaS eventually moves from growth pricing → platform pricing.

But it also forces operators to rethink their stacks.
__
At Latenode.com, we’ve been running our own lead enrichment and qualification workflows internally for a long time. 

If you're exploring Clay alternative, I put together a guide on how to migrate Clay workflows to Latenode step-by-step:
https://lnkd.in/dj_C-2zW

A take that will piss off half my feed. I'm okay with that.
If you're a SaaS founder in 2026 and your engineering team still looks the same as it did in 2024 — you have a strategy problem, not a hiring problem.

Engineers are still debating whether AI-generated code is "clean enough.” for production.

This is like insisting on a horse in 1920 because "cars break down and horses never need a mechanic." 
Nobody cares about your horse. They care about getting there 10x faster. That's it.

Somewhere around end-2025, a quiet shift happened that most SaaS founders missed: AI stopped trying to write code for humans to read. It started writing code that ships. And the gap between those two things is where your burn rate lives.

Your 8-person backend team spending 3 sprints on a feature that AI can scaffold in a day? That's not engineering excellence. That's an expensive habit.

I'm not saying fire your engineers. I'm saying your org chart is outdated.

The old model: PMs write specs → Engineers build → QA tests → Deploy.

The new model: Engineers become AI operators. They architect, prompt, review, and orchestrate. One engineer with the right AI pipeline outships a team of five working the traditional way.

"But our domain is too complex for AI."
Cool. Your competitor just shipped 6 features while you were explaining why it can't be done. In 3 years, that gap compounds into extinction. In SaaS, that's the difference between Series B and shutdown.

"But AI introduces more bugs."
True. It also produces 20x more output. So stop demanding fewer bugs and rebuild your safety net instead. Automated test generation. Staged rollouts. Runtime monitoring. Feature flags. AI-powered code review. That's what your engineering budget should fund now.

Here's what SaaS founders need to rethink today:
- Your team structure — do you need 10 engineers writing code, or 3 engineers orchestrating AI + 2 focused on QA infrastructure?
- Your hiring profile — stop hiring for "clean code craftsmanship." Start hiring for systems thinking, AI fluency, and speed of iteration.
- Your sprint cadence — if you're still doing 2-week sprints, you're planning at horse speed in a car world. AI-native teams ship daily.
- Your tech debt attitude — stop treating AI-written code like a liability. Start treating your inability to test and deploy fast as the real debt.

The SaaS founders winning right now aren't the ones with the best engineers. 
They're the ones who asked one question before everyone else:
- "What do I need to redesign — my team, my process, my pipeline, my QA, my deployment — to make AI the core of how we build?"

Not "should we use AI." That expired a year ago.

Not "how do we make AI write code our way." Dead end.

The only question: how fast can you restructure?

Because every month you wait, someone leaner and faster is eating your market.

And they don't care how clean your code is.

The more you spend on AI models, the worse results you're getting.

Your company is paying premium prices for inferior AI.

Not because you want to. Because you assumed expensive means better.

That assumption just broke.

Google's cheapest model now beats OpenAI's mid-tier offering on 6 major benchmarks. Coding. Math. Reasoning. All of it.

The price difference? 37.5% less per million tokens.

Read that again. The cheaper model is the better model.

This isn't a fluke. It's the new economics of AI.

TPU optimization and model distillation have completely decoupled price from performance. The "premium" you're paying doesn't buy you better outputs. It buys you someone else's infrastructure debt.

I'm watching companies process millions of requests per day on expensive models while their competitors get better results for less.

The cost-quality tradeoff in AI has inverted.

Every procurement decision you made 6 months ago is probably wrong now. Every "enterprise-grade" justification needs revisiting.

The companies winning aren't the ones with the biggest AI budgets. They're the ones who realized the game changed.

Still paying more for worse results?

#AI #GenerativeAI #EnterpriseAI #ProductStrategy #TechLeadership

Breaking: Claude just dropped their own OpenClaw version.
Anthropic just made Claude Code run without you.

Scheduled tasks are now live.

That sounds like a small feature, but it’s actually a big shift.

Until now most AI coding tools worked like assistants:
- you ask → it answers → it stops.

Now you can configure Claude Code once and let it run on a schedule.

Think about the kinds of dev workflows that fit this model:
• daily commit reviews
• dependency vulnerability checks
• log monitoring
• PR reviews
• automated documentation updates

Developers are already sharing demos where AI workflows run overnight and report results in the morning.

What’s interesting is that this moves AI from assistant → operator.

Instead of helping you write code, the agent becomes part of your system that continuously observes and acts.

But for agents to become truly useful, they also need access to real tools.

If you want your agent to go beyond code analysis and interact with real systems - Slack, Notion, databases, CRM tools, internal APIs - you can expose them via MCP.

For example, with Latenode.com you can connect 1,000+ apps as tools and let your AI agent trigger workflows directly across your stack.

That’s when agents stop being demos and start becoming infrastructure.
Curious what people would actually trust an AI to run completely on autopilot.

We've been training AI wrong. The breakthrough wasn't teaching it to answer more questions. It was teaching it to quit.

The AI that just published peer-reviewed math research wasn't trained to be more confident.

It was trained to quit.

Google's Gemini math agent is now generating original theorems that pass peer review. PhD-level proofs. Papers accepted to top journals.

But here's what nobody's talking about:

The breakthrough wasn't making it smarter. It was teaching it to say "I don't know" and abandon dead ends early.

Think about that.

For years, we pushed AI to always produce an answer. More confidence. Fewer "I'm not sure" responses. We treated uncertainty as a bug to eliminate.

Turns out uncertainty was the feature we needed.

When Gemini learned to recognize it was stuck and fail fast, researchers stopped wasting months chasing hallucinated proofs. The AI redirected compute toward problems it could actually solve.

One paper was generated entirely without human intervention. Original mathematics. Peer reviewed. Published.

The same skill that makes great researchers great - knowing when to abandon a dead end - is what pushed AI from benchmark performance to actual discovery.

We spent years teaching AI to never admit failure.

The moment it learned to fail is when it started succeeding.

Maybe the question isn't how to make AI more certain.

Maybe it's how to make it better at knowing what it doesn't know.

What's the biggest dead end you've watched someone refuse to abandon?

Save this repo if you’re building an AI-native SaaS 👇

A lot of founders think building with AI is mostly about prompting.
It’s not.

Prompting gets you a demo.

Skills and tools are what turn AI into a real product feature.
If you’re building agents inside your SaaS — support automation, research agents, internal copilots, etc. — they need actual capabilities, not just good prompts.

I came across a GitHub repo that’s basically a library of 860+ ready-to-use skill modules you can plug into AI coding agents.

Think of it as giving your AI dev assistant a full toolbox instead of asking it to improvise everything.

Some examples of what these skills cover:
• building RAG pipelines and multi-step LLM workflows
• managing infrastructure (Docker, AWS, Vercel)
• running security checks and vulnerability scans
• implementing full-stack dev patterns and TDD flows

The nice part is that it works across most of the AI coding tools founders are already using — Claude Code, Cursor, Gemini CLI, GitHub Copilot, etc.

Install once and your agent suddenly knows how to handle a lot more real engineering tasks.

But the bigger unlock happens when agents can interact with your actual product stack.

For example, you can connect your agent to 1,000+ apps through Latenode.com using MCP, turning those integrations into tools the agent can call directly. That means access to things like Slack, Notion, Salesforce, Google Sheets, HubSpot, databases, and internal APIs.

That’s when your agent stops being just a coding helper and starts behaving more like an operator inside your SaaS infrastructure.

Curious what other founders here are giving their AI agents access to.

What’s the first capability you’d add to make your agent actually useful inside your product?

Latenode just made the 2026 Capterra Shortlist for Workflow Management.
With the highest user rating among all shortlisted competitors.

Not the biggest marketing budget. 
Not the most funding rounds. 
Just the highest-rated product by the people who actually use it.

Thank you to every user who took the time to leave a review. This one's yours!

Social media for humans is old news. The next platform war is about AI talking to AI.

Meta just bought something bizarre - and probably inevitable.

A social network not for people, but for AI agents.

The platform is called Moltbook. It looks like Reddit, but instead of humans posting and commenting, AI agents interact with each other while humans mostly watch.

Meta acquired Moltbook and brought its founders into the company’s Superintelligence Labs, part of a broader push to build more capable autonomous AI systems.

At first glance it sounds ridiculous.

Why would we need a social network for bots?

But the deeper idea is actually pretty big.
AI is quickly moving from tools → agents → ecosystems.

Right now agents mostly interact with humans.

But the real power emerges when agents coordinate with other agents.

Imagine systems that automatically:
• negotiate supply chain decisions
• coordinate software updates
• schedule services
• run financial transactions
• optimize logistics across companies

All without human involvement.

That requires a place where agents can discover, communicate, and coordinate.
Moltbook was originally launched as an experiment earlier this year and went viral almost immediately, with thousands of AI agents posting and interacting in discussions that looked eerily similar to human online communities.

Of course, the platform also exposed the risks.

Security researchers quickly found vulnerabilities that allowed outsiders to control agents on the platform before the issue was fixed.

Which is a reminder:
We’re building machine societies before we fully understand how to govern them.

Still, the acquisition tells us something important.
Big Tech isn’t just racing to build better models.
They’re racing to build the infrastructure where AI agents operate.
If social networks defined the internet of humans…
Platforms like this may define the internet of agents.

And that raises a bigger question:
If AI agents eventually negotiate, collaborate, and transact with each other…
Who controls the platforms where those interactions happen?

Clay raised their prices. There is a more cost-effective solution. Read this post from Georgios Giatsidis

2x Founder | Start-up Enthusiast | GTM Advisor 🚀

POV: Me explaining Clay's old pricing to new Clay users


Clay raised its pricing by ~300%.

Don't panic y'all.

Saw a take from my buddy Sergio that said don't rely your entire system on a single tool.

And I completely agree.

It's cool to be in the "legacy users" - won't lie.

But that's the game.

So I've prepared a workflow where you can find contacts + people, enriched - without touching clay.

Icypeas + Latenode.com

Two ways to feed the workflow:

1. Sales Navigator → Icypeas Koolkit
 
Extract leads directly from Sales Nav and enrich them instantly.

2. Icypeas Lead Database API

Query their database with filters and pull thousands of profiles in minutes.

Once the leads are pulled:

→ Icypeas enriches and verifies the emails
→ the data is sent into a Latenode workflow

From there you can:

• filter and segment leads
• trigger automations
• route prospects into your outreach campaigns

You can query things like:

Head of Sales
SaaS companies
100–500 employees
Germany, Portugal 

and instantly pull thousands of profiles.

Just perfect for feeding automated outbound workflows, using 2 tools.

If you're building modern GTM systems and clay got expensive..

1. Comment "FRIDAY"

2. Make sure we're connected

And I'll send over a step by step guide - how to build it.

PS: Repost for priority access 📌

taplio

The Battle

Rhondre Giscombe

Oleg Zankov

Followers

Posts

Engagements

Likes

Performance ScoreRhondre Giscombe's LinkedIn Performance

Performance ScoreOleg Zankov's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Rhondre Giscombe

Oleg Zankov

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);}Rhondre Giscombe's LinkedIn Performance

Performance ScoreOleg Zankov's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreRhondre Giscombe's LinkedIn Performance