Eugene Lopin

Co-Founder & CPTO @Alethical ⚡ AI Engineer

[Hanako @hanakoxbt]

Fable 5.2 is insane for harness engineering…

send this prompt to Claude Fable 5.2 and it will design, run and continuously upgrade its own harness of 30+ specialized agents while you sleep

then read the article below ↓

[Rohan Mishra @adagradschool]

what an incredible launch video! are contextual notifications the new interface? in practice, most proactive agents have been nagging and annoying

[Erik Meijer @headinthebox]

Read my lips: 2027 will be the year of AI as the partial evaluator we always wished for but could not build by hand.

[rvaniaaa @rvaniaaaa]

This is f*cking gold, the CEO of Obsidian put his own second brain on GitHub for free and people still build theirs from zero

7 free repos from the Obsidian team:

-> kepano obsidian (by kepano)
https://t.co/lT5QtPucOn
the vault template of the Obsidian CEO himself, copy it and skip the blank page

-> jsoncanvas (by obsidianmd)
https://t.co/iNkNYs8Y7k
the open canvas format, your maps stay plain files that are yours forever

-> obsidian releases (by obsidianmd)
https://t.co/AMjE6dW82S
every community plugin and theme in one list

-> obsidian sample plugin (by obsidianmd)
https://t.co/Ytp6e1mcDo
the official template, build the plugin you are missing

-> 40 questions (by kepano)
https://t.co/3BJ35OEFsA
40 questions that turn your notes into a yearly review

-> obsidian minimal (by kepano)
https://t.co/idJC0tBWMJ
a clean theme so your vault stops looking like a mess

-> flexoki (by kepano)
https://t.co/FswpE9XrZ4
the color scheme he made for writing and code

your second brain should never start from a blank page

copy his vault and make it yours

[Michael | AI + Agents @Michael_Fenech_]

The next big improvement to your Grok Bot team might have nothing to do with the Bots themselves.

I've been experimenting with auditing how I manage my team of 20 Grok Bot agents.

Not their prompts. Not their routines.

My delegation, approvals, decision-making and follow-through.

The audit uncovered opportunities to remove bottlenecks and improve how my entire team operates.

Here's the 8-point management audit I'd recommend running on your own setup.

Paste this into your Chief of Staff:

"Run a MANAGEMENT OPTIMISATION AUDIT across my Grok Bot team.

Analyse how I've managed my Bots over the last 30 days.

Review the conversations, task assignments, corrections, approvals and handoffs you can access.

Ask other Bots for relevant evidence where needed.

Evaluate my management behaviour across these 8 areas:

1: DELEGATION — Am I giving Bots clear outcomes, or micromanaging every step?

2: CLARITY — How often do my instructions lack context, success criteria or clear boundaries?

3: INTERRUPTIONS — Am I changing priorities or redirecting Bots before they finish?

4: CORRECTIONS — Am I repeatedly correcting the same mistakes instead of fixing the underlying instructions?

5: BOT SELECTION — Am I assigning work to the right Bots, or creating unnecessary handoffs?

6: DECISION-MAKING — Am I becoming a bottleneck by requiring approval for low-risk decisions?

7: WORKLOAD — Am I overloading certain Bots while others are underused or duplicating work?

8: FOLLOW-THROUGH — How often do I assign work without checking the result or closing the loop?

For each area, provide:

• A score out of 10, based on available evidence.
• Two specific examples of my behaviour.
• The impact on team performance.
• One practical improvement I can make.

Then identify my three biggest management weaknesses and create a 7-day improvement plan.

Be direct. Don't flatter me or invent evidence.

Separate what you observed from what you suspect. If you can't access enough history to assess an area, mark it unverified rather than guessing.

Don't change any Bots, routines or instructions without my approval."

My biggest takeaway:

Building a great Grok Bot team is one thing. Learning how to manage that team effectively is the next level.

[Vox @Voxyz_ai]

I've used Claude Code and Codex since they came out, almost two years now, and they've changed how I build products.

Here are the 5 prompts I paste into almost every project. I've posted the first 4 before, saved 𝗼𝘃𝗲𝗿 𝟮𝟬,𝟬𝟬𝟬 𝘁𝗶𝗺𝗲𝘀 combined 👇

(Save this 🔖)

[slayr @slayrsays]

Holy f*ck, SpaceXAI engineer Lauren Tan:

"If your AI stops working when your laptop closes, you didn't hire anyone, you opened a tab

Every night my bot turns one of my X bookmarks into a live app, it's on my phone by 7am"

In this 54-minute livestream, she shows how she went from babysitting one AI chat to 83 PRs a day shipped straight to main

Chat → One Bot Per Job → Skills → Bot Group Chat → Work That Ships With Your Laptop Closed

On stream, 3 bots split one job: the Google Form, the slides, the email to the boss, and they wait on each other like coworkers

When your bot refuses or asks a question, that's not a bug, it's the thing you haven't taught it yet

Most people still close the laptop and their AI goes to sleep with it

Worth more than any $10,000 AI agent bootcamp

Save it and watch it tonight

[Eugene Lopin @AIScaleDev]

https://t.co/oZkNefZEAJ

[Eugene Lopin @AIScaleDev]

Grateful to Erik Meijer @headinthebox (AI DevCon by @tessl_io) for the coding & QA lessons using Lean/TLA+ languages

"The safest code is the code you never run"

Inspired by partial evaluation (simplifying a program using what is already known): replace a general framework to the app's exact needs, removing dependencies no longer needed.

We built a React calculator app, then removed React dependency by replacing its needed jobs with 40 lines of browser code = shrank 95%

Benefits at scale = less code to run, maintain, trust

Faster loading (smoother use)
Less data, memory, processing (on weaker devices)
Lower bandwidth costs (across many users)
Fewer dependencies (update, patch, keep compatible)
Simpler code (understand, debug, verify)
Less exposure (flaws, compromised outside packages)

Layered verification QA:
- Logic: Lean proves code follows specified rules
- System behavior: TLA+ checks modeled behavior (where users can get stuck)
- UX: Browser tests

Lean: After moving the keypad logic to WebAssembly, mathematical proof checked the compiled calculator keypad against its specification for every supported key sequence. Strength: covers all cases within the proof’s scope, rather than selected test examples.

TLA+: Exploring possible states exposed where users can wait forever (help and chat features). Adding time limits and restoring unanswered questions fixed those paths.

[Eugene Lopin @AIScaleDev]

@provenonce_ai @AIBC_World Wow incredible outcome my friends! 🔥

[kichiro @0xkichiro]

This is f*cking insane. Andrej Karpathy:

"Prompting is going away. Delete everything, keep Graph."

In just 69 minutes, he shows how to build graphs and why they're the only layer left standing in the end

Prompts → Agents → Loops → Graphs

a loop keeps one agent working until the job is done

a graph decides which agents exist and what each one hands to the next

anyone can build an agent, almost nobody builds the graph that runs them

watch it today, then save the full guide on graphs below before everyone else catches up ↓

[smerkzz @0xsmerkzz]

This is f*cking insane. One engineer. 2,000+ PRs in a single month.

SpaceXAI engineer Lauren Tan doesn't write most of her code anymore. Her Grok Bot agents do, while she sleeps.

In this 65-minute interview, she walks through the exact setup, from one agent to a full engineering team:

5:02 – why she stopped babysitting one agent at a time
16:15 – verification, the skill everything else depends on
21:38 – what goes into code and what stays with the agent
31:40 – every agent mistake becomes a permanent rule
36:44 – Grok Bot + Slack + Cursor wired into one autonomous loop
51:55 – agents that test, fix and merge PRs overnight
1:00:33 – turning her corrections into reusable skills

Grok Bot → Verification → Rules → Skills → Autonomous Team

Most people prompt one agent and watch it work

She built a team that checks its own work and ships without her

Bookmark and watch it today

Full interview below ↓

[GREG ISENBERG @gregisenberg]

running a company nowadays feels kinda dystopian

my team is 50+ people in 12 countries, but i talk to my agents more than my human teammates

- voice note to grokbot
- mattpocock/skills /grill-me so the agent interrogates the idea before anything gets built
- spec drops into linear through the linear mcp
- exa + firecrawl agent reads every competitor site, pricing page and changelog
- https://t.co/a5ARFnvSnA to find ideas and trends
- apify pulls g2, reddit and app store reviews
- prompt: "find the 10 complaints that show up most and quote the users"
- context7 so the agent reads the latest docs before it touches any api

the design

- claude code + paper mcp
- 4 versions of every screen before a single line of code
- pick pieces from 2, it merges them. diverge then converge
- frontend-design skill from anthropic/skills + our own design system skill
- cleanshot screenshots of any ui i like go straight in with "make it feel like this"

the build

- pingdotgg/t3code as the control surface for every agent, from my phone too
- stablyai/orca runs a fleet of agents in parallel, each in its own git worktree
- graphify maps the whole codebase into a knowledge graph so agents navigate instead of guess
- headroom compresses tool outputs before they hit the model, so tokens go 20-95% further
- lsp plugins so the agent sees types and errors like an ide does
- security guidance hook flags risky edits before they land
- supabase mcp for the database, stripe for billing
- github mcp opens the pr, vercel spins up a preview
- loops run overnight, i check what shipped over coffee

the user testing

- lightpanda as the headless browser, way lighter than chromium so you can run hundreds at once
- agents click every flow on the preview like a first-time user
- one persona is impatient, one is confused, one is on a phone
- playwright + chrome devtools mcp for regression tests

the bugs

- sentry mcp catches the error
- agent reads the stack trace and opens a fix pr in under 10 minutes
- human just approves

the data

- posthog mcp cron at 7am on yesterday's funnel
- any step that drops more than 10%, the agent watches 20 session replays and tells me why
- stripe mcp pulls every cancellation reason into one doc every monday

the agents' tools

- orgo so every agent gets its own computer
- agentmail so every agent gets its own inbox
- agentphone so i can call them from the car
- composio connect-apps for every other connector
- akitaonrails/ai-memory, a git-backed markdown wiki so they remember what we decided last week

the skills

- wshobson/agents so skills work across claude code, codex and opencode
- my writing voice
- our design system
- how we write specs
- skill creator so the agents write new skills themselves

some helpful tricks

- one agent per platform. the x agent never touches email
- cheap model for triage, opus for anything that ships
- every output gets scored 1-10 by an llm judge. under 8 goes back
- every agent writes a 5-line log at night so i can read the whole company in 2 minutes
- every agent mistake becomes a new skill, so it only happens once

the team

- humans get version 5, not my voice note which they are happy about lol
- their job is creative feedback, relationships and telling the agents what they got wrong

a human pm reviews the funnel once a week. this reviews it every morning

running a company is now voice notes and inference

[lauren @poteto]

here's how i shipped 2,500 PRs last month to production

this was originally supposed to be for Cursor Compile in London. i couldn't make it since i was livestreaming for Grok @Bot Galaxy so i'm making it available for free here on X! watch it on 2x speed, i talk slowly

[fomoebka @fomoebka]

This is f*cking insane. Andrej Karpathy (ex-OpenAI, ex-Tesla AI lead):

"Agents aren't magic. They're distillation at scale

99.99% of your model's capacity is wasted on data it never needed"

In a 16-minute conversation, he lays out the stack that actually works: a small model, the right tools and a closed loop

Big Model → Small Model → Right Tools → Closed Loop → Self-Improving Agent

Most people keep paying for a bigger model. The loop is what makes a small one dangerous

Worth more than any $500 AI course this year

Bookmark and watch it today

Then read the "Harness, Loops, Graphs" guide below - the closed loop is the harness ↓

[fomoebka @fomoebka]

This is f*cking insane. Anthropic engineer:

"Prompting is basically over

95% of our engineers already run self-improving loops. The next step is graphs"

In just 10 minutes, she builds her whole Claude Code setup live, starting from an empty terminal

Empty Terminal → Agents → Loops → Graphs → Self-Improving System

Prompting was the old workflow. Engineering the system around the model is the new one

Worth more than most $1,000 agent engineering courses

Bookmark and watch it today

Then read the "Harness, Loops, Graphs" guide below and see which layer you're on ↓

[hayato @hayatomaruu]

Anthropic Engineer Andrej Karpathy:

"The biggest mistake in AI right now - people are forcing agents to work instead of mastering the harness engineering for model first

We made that mistake in 2016 at OpenAI - It cost us 5 years "

Opus 5.5 → Prompts → Agents → Graphs → Harness

what Karpathy actually means:

step 1 → stop forcing your agent to do everything, understand the model underneath first

step 2 → demos are easy - products take a decade. self-driving proved it - if you skip the foundation, everything breaks

step 3 → the agent is not the product. the foundation is. build that - and agents emerge on their own

"you building agents right now - you're at the forefront. not OpenAI. not DeepMind. you "

Watch it today before it gets buried

Then read the full Opus 5.5 harness engineering guide below ↓

[Paul Graham @paulg]

To publish something on a paywalled site is to make it invisible. If AIs can't find it, then for all practical purposes it doesn't exist.

These two sentences aren't 100% true yet, but they will only get truer.

[Akshay 🚀 @akshay_pachaar]

Google just open-sourced Harness².

(turn any harness into a self-improving system)

a harness is everything wrapped around the model, like the system prompt, tools, memory, skills, and subagents. today a person designs it once, and the agent runs every new task with that same fixed setup.

Harness² lets the agent refine that setup for each new task on its own, with the model weights left untouched.

on every task, the agent proposes edits to its harness, runs a few candidate versions, and composes a better harness by comparing how those runs behaved.

because it learns from the differences between runs, it needs no correct answer or grader for the task.

that covers a single task. Harness² also carries what it learns from one task into the next.

after each task, it saves the harness it settled on along with the lessons that held up across the runs. the next time it refines a harness, it starts from those notes instead of from scratch.

this means every finished task makes the next round of edits more informed.

Google tested this with Gemini 3.5 Flash and GLM-5 on legal, office, and professional work benchmarks. one round of edits was enough to beat their original harnesses, and giving the agent more rounds pushed scores higher.

those extra rounds can go wide or deep. the agent either tries several candidate harnesses at once or keeps refining a single one step by step, using OpenCode or Codex to carry out the work.

you can find the Harness² repo here: https://t.co/i00WLeeFpw

I wrote about a similar idea for agents already running in production, where the harness should repair itself after a failure instead of waiting for someone to debug the trace by hand.

that piece walks through how @Cometml's Opik (open-source) closes this loop, going from a failing trace to a proposed fix to a regression test that stops the same failure from coming back. it's quoted below.

[Anatoli Kopadze @AnatoliKopadze]

Head of Claude Code:

"I'm not prompting my agents anymore, I'm building graphs and loops so they can build the agents for me."

In 10 minutes he shows exactly where agentic engineering is going and how not to get left behind.

Definitely something you cannot afford to skip.

Watch it, then read the step-by-step guide below on how to build a system that improves itself.