Qweexee

That girl who opened AI "for five minutes" and is now building entire projects with it.

[Qweexee @Qweex_eth]

SpaceXAI engineer, Lauren Tan:

“Right now, I’m running 30+ agents in a loop with /pstack. I shipped 1,000 PRs last month, this month, I’ll 2x that.

I woke up today and 20 had already landed. My job right now is to build a team of agents.”

in 1-hour session, SpaceXAI engineer explained how to build an effective team of AI agents that work 24/7

it’s worth more than any $500 agentic engineering course.

watch it today, then read how to build the same agent fleet in one click in the article below ↓

[Qweexee @Qweex_eth]

Cursor engineer, Lauren Tan:

"I woke up. 20 PRs had already landed on main. I reviewed them. They were good. I shipped 1,000 PRs last month. If your agent can't verify its own work, you are the verifier. And you're the bottleneck."

99% of engineers still run a single chat window. That's not a skill gap. That's a stack gap, and it takes one evening to close.

/pstack → Verification → Skills → Evals → Cloud agents → Auto-merged PRs

In a 1-hour session, a Cursor engineer explains how to build an effective team of AI agents.

Research → Build → Launch → Improve

It's worth more than any $1000 agentic engineering course.

[Qweexee @Qweex_eth]

Anthropic engineer:

"Using Claude efficiently doesn't mean using fewer tokens. It means the tokens you do use go toward what you actually asked.

One prompt is many requests, and every request carries the entire conversation. The cache keeps that cheap, as long as you don't break it"

In a 25-minute workshop, an Anthropic engineer showed where Claude Code tokens actually go: cache, model choice, effort level, context, subagents

this is worth more than a $500 vibe-coding course

watch this workshop today, then pick your model, effort and MCP servers once at the start of a session and leave them alone

[Qweexee @Qweex_eth]

Fair point, and the talk doesn't cover it. He only says he reviews the landed PRs on main after the fact. What he does stress is that the gate has to be real verification: the agent runs the app and checks its own work, and the skills behind it are tested with evals.

Your question is about provenance, which is a separate layer. A few things would answer it:
→ run checks in a merge queue against the actual base, so the tested tree is the one that lands
→ record the commit SHA and tree hash next to the verification result
→ keep the evidence (traces, screenshots, eval scores) as durable artifacts attached to the commit, not just in CI logs that expire
Without that, "it was good at merge time" is a claim you can't prove later.

[Qweexee @Qweex_eth]

Cursor engineer, Lauren Tan:

"Most people use agents while still checking every output by hand. If your agent can't run and verify its own code, you are the bottleneck.

I started with one agent I watched constantly. Now my agents auto-merge PRs: last month I landed 1,000, and this morning I woke up to 20 already on main"

In a 1-hour session, a Cursor engineer showed how to go from babysitting one agent to trusting a fleet of them: verification, skills, evals, cloud agents

this is worth more than a $500 agentic engineering course

watch this workshop today, then build your own verification skills with the Pstack plugin

[Qweexee @Qweex_eth]

@axnatuh The tech is easy; building the distribution engine is where the real leverage is. Great breakdown.

[Qweexee @Qweex_eth]

@sopersone Funds are in, 10 tentacles are crawling, and the product actually prints. $10k locked, GitHub dropping tomorrow—we’re moving fast!

[Qweexee @Qweex_eth]

@sopersone $SQUID is unstoppable! 10 tentacles pushing code and building the future while the market sleeps. Can’t wait for the GitHub release!

[Qweexee @Qweex_eth]

A friend told me to watch this talk. It replaced a $700 course, and I've already started applying it to paid work.
Delete your IDE. Stop building one agent and start building a team of them.

In a 24-minute talk, Steve Yegge and Gene Kim lay out how AI engineering will look from here on. Their claim: all code will soon be written by agents, while engineers oversee the work without reading every line.

Steve's hot take: if you are still living in an IDE after January 1st, you are a bad engineer. Gene backs it up with real case studies and a simple finding: the longer people use AI, the more they trust it.

The big idea: stop building a bigger diver. Build a team of agents, each with one job.

[Qweexee @Qweex_eth]

@axnatuh Wooow man!! Ty for support

[Qweexee @Qweex_eth]

@sheemamoto This is the real shift in AI - not one powerful agent, but an entire team of agents working together.

[Qweexee @Qweex_eth]

@SamiKowalski This is actually so good

[Qweexee @Qweex_eth]

SpaceX AI engineer Nick Miller: "Bots can really function as a team of knowledge workers. I have a travel bot that talks to a flight bot and a hotel bot, and they book everything for me. I have a closer bot that finds every promise I made in a customer meeting and drafts the follow-up emails."

GrokBot is the fastest-growing product its team has ever built: 418,000 weekly users.

In a 27-minute demo, a SpaceX AI engineer shows how to build a team of GrokBot agents that work together:

— One bot found, scheduled viewings for, and negotiated a Singapore rental lease from start to finish.

— Another bot renegotiated a contract and saved tens of thousands of dollars.

— Single engineers are now shipping with the impact of 10.

— Every bot gets its own cloud computer, so it keeps working after you close your laptop.

Key moments from the video

4:25 — The Singapore housing story: the bot researched ~35 rentals, built a Google Sheet, emailed his wife, booked viewings and negotiated the lease.

8:35 — How to build a bot: Context, Connectors, Outcomes.

10:24 — Bots persist and learn from your corrections, instead of starting a new session every time.

11:47 — Shared cloud computer with its own browser: no API or MCP needed.

15:03 — Executive use cases: the 5 a.m. morning briefing and meeting prep from Slack, Gmail, Notion and call transcripts.

16:55 — Engineering: a team of bots (front-end, QA, docs, research) that builds through cloud agents.

19:49 — Bots talk to each other: the travel bot delegates to the flight bot and the hotel bot.

22:41 — Sharing and the bot marketplace, plus the "closer bot" that turns meeting promises into draft emails.

24:12 — HaggleBot: renegotiated a Zoom contract and saved tens of thousands.

24:36 — His 4 rules for a strong foundation: give access, teach your bots, let them create skills, delegate between bots.

Worth more than a $500 agentic engineering course.

Skip Netflix today and watch this.👇

[Qweexee @Qweex_eth]

@SamiKowalski Amazing how brief videos like this changes our vision in AI usage in our everyday lives.

[Qweexee @Qweex_eth]

Anthropic, Head of Design:

"Claude Code is the most powerful tool for shipping products, but 99% of people use it wrong.

to get tier 1 results with Claude Code, you should run parallel sessions in worktrees, stay in auto mode, build a /prototype skill & let Claude open the PR with a screen recording - that's the new stack of a designer who ships."

in 12-minute stage, Anthropic's head of design gives main tips for using Claude at 100% of its power.

worth more than a $500 designe engineering course.

watch this video, then read the full guide on Claude Code workflows in the article below.

[sopersone @sopersone]

https://t.co/TqxlmoZhvl