Lucas Ribeiro

You can't do everything, but you can do anything.

I like big data and I can't lie.

[Lucas Ribeiro @LRiverside]

“… Being useful - leaves a trail”

- Sao Josemaria escrivá

[Lucas Ribeiro @LRiverside]

@Pallavi_345 But definitely time for a new

[Lucas Ribeiro @LRiverside]

@Pallavi_345 Water and microfiber , light detergent at most

[Lucas Ribeiro @LRiverside]

#dreaminindata Chicago ‘26

[Ptah.ai 🇯🇲🇵🇷🏴‍☠️ @ptahdunbar]

Shipped my first contribution to OpenClaw today, landed in the 2026.4.14 release ✨🦞✨

If you're building agentic systems, gon' ahead and upgrade 🚀

Thanks @obviyus, @vincent_koc and the OpenClaw maintainers for the quick review.

Godspeed fellow #clawtributors 🏴‍☠️

[Lucas Ribeiro @LRiverside]

@abhitwt @VibhuRevadi twitter. you should definitely install twitter.

[Lucas Ribeiro @LRiverside]

@L1AD @chamath @tobi I think this a tool in the agents kit but not the solution

[Lucas Ribeiro @LRiverside]

@Invest_Brandon @chamath There’s a persistent memory solution paper that does this better without updating md.

That would ruin the context regardless of budget.

[Lucas Ribeiro @LRiverside]

@chamath @Invest_Brandon You have a connector that streams data to you lake that is being queried.

[Lucas Ribeiro @LRiverside]

@chamath a paper recently on peristent memory systems for ai to keep context.

3 agents that’s primary jobs are to keep the separate forms of context.

1 agent that interprets ?s and sends req formatted to another query agent that looks over all , and sends back and checker

[OpenClaw🦞 @openclaw]

OpenClaw 2026.3.2 🦞

💬 Telegram live streaming
🔌 ACP subagents on by default
📄 Native PDF tool
✅ openclaw config validate
🇻🇳 Zalo rebuilt in pure JS
🔒 100+ security & stability fixes

Sleep is a feature we haven't shipped yet.
https://t.co/4EDqZBHcBc

[Lucas Ribeiro @LRiverside]

King Trumps Dictator ?

Makes sense in D&D, makes sense in life.

[Lucas Ribeiro @LRiverside]

@supabase wtf is happening

Severe data loss before weekend before we have investor pitches, need help ASAP

https://t.co/HTaRoXwO8h

[Theo - t3.gg @theo]

Meta just dropped a new code model. Goal isn't for us to use it as devs - it's to demonstrate "code world modeling" as a training method.

Fascinating. Benchmark performance is solid too!

[Lucas Ribeiro @LRiverside]

@JoClarkEntrprnr @Jason He’s opening your mind. Conversation on the way. LOL

[Lucas Ribeiro @LRiverside]

@Jason …. Just building tools to make days data engineering easier.

[Lucas Ribeiro @LRiverside]

@johnmaeda @joulee Great episode

[Lucas Ribeiro @LRiverside]

This tweet is the formula / roadmap for the delta of “future truth” vs talking shit.

Promises can be made, but unless this is central to your process, there’s a big fork leading to drastically divergent destinations.

[Julie Zhuo @joulee]

Diagnose with data. Treat with design.

One is the yin and the other the yang. They follow each other in an infinite cycle.

Data is the process of deeply understanding reality as it *is*.

Design is the process of designing future reality as it *could be*.

[Grant Lee @thisisgrantlee]

we grew from zero to $50M ARR in <2 years, profitably
i've never publicly shared our tactics before
it's cost us over $5M to learn what I'm about to share

800-word long post on every growth hack that printed money for us

I'll cover:
1. Influencer Marketing
2. Performance Marketing vs Brand
3. User Testing
4. Dogfooding

0-10M was nearly 100% word of mouth and organic content

10-50M was still >50% word of mouth but influencer, affiliate, and referral made up the other half

But the foundation we set from 0-$10M is what allowed us to blaze past $50M

1. Influencer Marketing 101:

90% of your reach comes from <10% of content that goes viral. Your job: go broad with influencers and spend enough to find that 10% of content and formats that work. Virality is no accident - test and discover hooks, visuals, formats for each platform, understand why they work, and replicate 100x across your influencer roster.

Most startups get three things wrong:
1 - Too small budget
2 - Overly selective on creators and messaging
3 - Give up too early
Start with $10-20k/month, commit to 6 months minimum. This lets you experiment with many micro influencers and lets you test many concepts.

- List creator personas with audiences that care about your product. Be exhaustive. Influencer outreach is time-consuming. Work with a freelancer or agency if you don't have a team. DM me for recommendations.

- Offer base + viral bonus. For TikTok, have them start new accounts. New accounts can get as many views as established ones. But on new accounts creators are unburdened by their brand and can try any format.

- Test across all platforms (TikTok, IG, X, LinkedIn work for most B2B/prosumer apps).

- Track specific creators and hooks that outperform, not just channels.

- Add 'How did you hear about us' to onboarding. Calculate which channels drive leads vs just views.

- Compile every asset, video or hook that works into a guide. Once you have 20-30 winning formats, hire your top creators as consultants to train all other influencers on what angles go viral for you and why.

- Never write scripts - it becomes an ad no one watches.
Creators know their audience best.

2. Invest in brand before performance marketing

Most startups are not prepared to scale their spend because their brand and creative sucks. We had to go through an extremely expensive rebrand (both in terms of time and dollars spent) to land on something we loved. But it would have been much more expensive to skip this step.

Be relentless about testing a wide range of creatives. Whatever number of creatives you're thinking of testing, 10x it. Be bold. See what moves the needle (CAC payback, conversion rates, LTV and retention).

Once you see a particular use case resonating, build a whole funnel around it with symmetric messaging. You don’t want to confuse visitors by showing them an ad with one value prop and then drop them on to a landing page with a completely different message.

3. Get users to test prototypes before you ship.

This seems trivial, but saves you from the worst startup mistake - misdirected product development.

Most startups think experimentation is for mature companies improving their funnel by 1% through a/b testing. But Gamma's founding team is early Optimizely folks. Experimentation is in our DNA.

We ran extensive user testing on platforms like voicepanel and usertesting

a. Recruited random people who make slide decks for work. They have no skin in the game and won't lie to be nice.

b. Showed them prototypes with minimal instructions and let them spend a few minutes trying things out.

c. Asked them to think out loud. The words they use to navigate, their expectations, where they get confused is gold. Combining their voice + screen actions amplifies things 10x. You really feel the pain of poor UX and overcomplicated copy.

Did this for everything - landing pages, onboarding, new features, moonshot concepts.

Always test your biggest assumptions about what people will get or like. Identify major blind spots in your prototype stage, not at the end.

Ship once you have hours of testing proof that ordinary people find it easy and useful.

4. Dogfood the hell out of your product

Either you build something that's 100x better than alternatives, or you build something else. Dogfooding makes it painfully obvious if your product isn't a 100x better.

Storytime: When we started Gamma, we had two competing ideas -
Idea 1: a virtual office to capture in-person magic
Idea 2: reimagining PowerPoint from the ground up (This became Gamma)

After 6 months of building and dogfooding both of them in parallel, we had a clear winner.

It wasn't about metrics or hard data.

With a virtual office, we were always competing against IRL and could never replace the magic of being together in person.

With the reimagined PowerPoint, we could imagine a product 100x better than the alternative.

Any startup can learn from our playbook.
I really hope they do.

[Lucas Ribeiro @LRiverside]

@krishdotdev Plead the 5th