Founding growth lead at a FinTech startup @joinkudos. Passionate about data driven product and marketing strategy. Always a student.
Sunnyvale
Joined February 2021
- Tweets82
- Following89
- Followers25
- Likes57
Pinned Tweet
Google invented the Transformer and employed every AI pioneer that matters. They even had ChatGPT-level tech ready in 2021.
Yet OpenAI's launch caught them "completely flat-footed."
How the greatest business in history wound up playing catch-up 👇 @AcquiredFM
I've been living in Claude Code for months, so I went into Cowork half-expecting a watered-down version. It genuinely surprised me.
Some things it just does better: scheduled tasks with zero setup (a Gmail + Slack daily brief that took me an hour to wire up in Code took 10 minutes in Cowork, no repo, no keys), and one-shotting clean docs and dashboards to hand off.
I ran the same 5 tasks through both to map out when to reach for which. Full writeup with the token numbers: mearaalgama.substack.com/p/c…
Where have you found Cowork actually beats Code?
"Cowork is just Claude Code with a nicer UI."
It's technically true and completely useless.
After a week of running the same 5 tasks through both, here's the answer that actually helps.
The framework I landed on:
→ Reach for Cowork when you need less control, the scope is small, and the connectors already exist.
→ Reach for Claude Code when you'll iterate on it, version it, or run it as a system that should get smarter over time.
A few things that surprised me:
- On a daily brief, Cowork burned only ~20% more tokens than Code. Basically a wash.
- Writing to a Google Sheet, Cowork used 350% more, because with no Sheets API it drives the browser by hand: 160 keystrokes and 5 screenshots to write 10 rows.
I'm not an engineer, and I still built real autonomous systems in Code. The "Code is for developers" line is a myth.
Full breakdown with the decision framework in first reply
The thing people don't realize is that utility companies already use AI against you. Automated systems designed to loop you, exhaust you, and keep you paying more. We just leveled the playing field
Mea Algama retweeted
Our newly launched @joinkudos AI agent called T-Mobile and cut my bill by $840/year (54%).
It handled the hold music, the bots, and the negotiation.
Companies have call center AI agents. Now consumers have theirs too.
Here’s the clip of the call.
Absolutely delightful episode. Most surprising learnings: why Joe Coulombe sold TJ (income taxed at 73%!?), how TJ/Pronto started as a 7-11 knockoff, and how Joe was responsible for popularizing Napa Valley wines.
Highly recommend listening! @AcquiredFM
Episode here: acquired.fm/episodes/trader-…
Careful: complexity early is often information debt. If I simplify too soon, I hide real constraints. Earn simplicity after you map them.
This Post is from a suspended account. Learn more
Mea Algama retweeted
"The quantity of AI-generated articles has surpassed the quantity of human-written articles being published on the internet"
March 2023 — Google rushes out Bard
Google launches Bard (now Gemini) just 4 months later.
It's seen as rushed, incomplete, playing defense.
Microsoft integrates ChatGPT into Bing. Google scrambles to protect Search.
The $140B profit machine they couldn't risk disrupting? Now under direct threat.
December 2022 — Google's "Code Red"
ChatGPT's explosion triggers panic at Google.
Sundar Pichai declares "code red." Teams get reassigned overnight.
Larry Page and Sergey Brin—who stepped down in 2019—attend emergency meetings for the FIRST TIME IN YEARS.
The company that invented modern AI is now playing catch-up.
November 2022 — ChatGPT drops
OpenAI releases ChatGPT to the public.
It goes VIRAL overnight. 100M users in 2 months.
The kicker? Google already had similar tech (LaMDA) ready in 2021.
But they didn't release it due to "reputational risk."
Too big to move fast. Too much to lose.
2019 — Microsoft × OpenAI
OpenAI restructures (capped-profit model).
Microsoft invests $1B + gives them unlimited Azure compute.
The deal: OpenAI gets the resources to train massive models. Microsoft gets the best AI in the world for Azure.
Google's talent is now building against them.
2018 — OpenAI's move
Google treated Transformers as research.
OpenAI treated them as a PRODUCT.
GPT-1 drops: the pretrain-then-adapt playbook that changes everything.
While Google publishes papers, OpenAI is shipping.
2017 — "Attention Is All You Need"
Google Brain publishes the Transformer paper—a clean break from RNNs/CNNs that makes large-scale pretraining actually practical.
They give it to the world for free.
OpenAI is watching closely.
2015 — The Rosewood dinner (the betrayal)
Elon was furious Google bought DeepMind.
So he and Sam Altman invite Google's best researchers to dinner with a pitch:
"Leave the trillion-dollar company. Build AI for humanity, not shareholders."
Reid Hoffman & Peter Thiel write checks.
2014 — Google buys DeepMind
Now Alphabet has EVERYONE: Hassabis, Legg, Suleyman, plus the AlexNet trio.
They're not just doing research—DeepMind would go on to cut Google's data center cooling costs by 40%.
This should've been game over for everyone else.