Relivs (@Rel1vs337) header
Relivs (@Rel1vs337) profile picture

Researching Frontier Building Agentic Setups

AI/acc
Joined December 2024
Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)Photo by Relivs (@Rel1vs337)
Relivs retweeted
Holy shit, this is absolutely insane... Anthropic engineer: "At Anthropic we don't write prompts anymore. We build loops and graphs." In 32 minutes she shows how the Claude team builds systems that prompt themselves Prompts → Agents → Loops → Graphs skip a layer and it comes back as a failure you blame on the model by the time you find it the week is already gone and the layer people skip most is the one where the shape gets decided this beats any $1,000-$2,000 course on agents I've come across watch it today, then save the full graph engineering guide below ↓
4
41
1,146
Holy shit, this is absolutely insane... Anthropic engineer: "At Anthropic we don't write prompts anymore. We build loops and graphs." In 32 minutes she shows how the Claude team builds systems that prompt themselves Prompts → Agents → Loops → Graphs skip a layer and it comes back as a failure you blame on the model by the time you find it the week is already gone and the layer people skip most is the one where the shape gets decided this beats any $1,000-$2,000 course on agents I've come across watch it today, then save the full graph engineering guide below ↓
4
41
1,146
Fuck, this is absolutely insane... 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 really means: step 1 → stop delegating everything to the agent. understand the model underneath it first step 2 → a demo takes a weekend. a real product takes years. self-driving showed that - cut the base layer and the whole stack collapses step 3 → the agent is not the thing you're building. the foundation is. get that right - and agents follow naturally "you building agents right now - you're at the forefront. not OpenAI. not DeepMind. you " Worth more than any $3,500 agentic engineering course Watch it today before it gets buried Then read the full Opus 5.5 ai engineering guide below ↓
5
31
2
313
32,010
This is f*cking insane. 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 really means: step 1 → stop offloading everything to your agent, master the model underneath it first step 2 → demos come fast - products take a decade. self-driving showed it - cut the foundation and everything falls apart step 3 → the agent is not the product. the foundation is. build that first - agents follow on their own "you building agents right now - you're at the forefront. not OpenAI. not DeepMind. you " Bookmark it and watch before it gets buried in your feed Then read the full Opus 5.5 harness engineering guide below ↓
5
21
4
219
26,162
This is f*king insane. Andrej Karpathy spent 8 years at OpenAI and Tesla Last week, he packed everything he learned into one free 2-hour lecture Agents → Harness → Loops → Graphs 95% of people still use AI like a calculator - this is what the other 5% already know You might not have 2 hours right now Bookmark it and watch before it disappears from your feed Then save the full agent harness guide below ↓
2
2
28
912
This is f*cking insane. Andrej Karpathy: "Prompting is starting to fade Leave the old stack behind - keep Graphs" In a 1-hour talk, he breaks down how to build Graphs and why they'll outlast everything else LLMs → Prompts → Agents → Graphs A prompt gives you one answer, then drops the task entirely An agent finishes the job, but forgets everything it learned A graph makes every finished job the starting point for the next Watch it today Then save the full guide to Graphs below before everyone catches up ↓
26
833
This is f*cking insane. Dario Amodei spent 6 years at Google Brain and OpenAI, where he led the team that built GPT-3 Then he left to start Anthropic, and now he's sitting on a model too dangerous to release In this 1-hour interview, he breaks down why Mythos stays locked even though it could print money tomorrow "We didn't plan for 80x growth. It would not have been rational" Mythos already found 271 new Firefox vulnerabilities, plus thousands more in private codebases Early testers called it a super weapon. One said you should need a gun license to use it Anthropic bleeds money every single day it stays unreleased GPT-3 → Claude → Opus 5.5 → Mythos 95% of people still use AI like a calculator - the 5% who understand what Mythos is will run everything You probably don't have an hour right now Bookmark it before it disappears from your feed Then read the ai engineering guide below and get ready before Mythos ships ↓
1
1
19
858
This is f*cking insane. Sam Altman (CEO of OpenAI): "You don't need to write prompts anymore. You need loops and graphs that write them for you." In 79 minutes he breaks down how to use LLMs better than almost anyone does Prompts → Agents → Loops → Graphs miss a layer and it comes back as a failure you pin on the model by the time you catch it the week is already gone people spend $15k on bootcamps that teach less than this watch it today, then save the full graph engineering guide below ↓
1
3
25
592
This is f*cking insane. Anthropic engineer: "You're not supposed to prompt Opus 5.5 You're supposed to build a system that prompts itself. At Anthropic, 90% of our engineers have already built such systems" In 30 minutes she breaks down exactly how Anthropic builds agents that remember, fix their own errors and get sharper with every run. Worth more than any $1000-$2000 agent course out there. Watch it today, then read the article below ↓
2
1
24
724
This is f*cking insane. Andrej Karpathy spent 8 years at OpenAI and Tesla Last week, he packed everything he learned into one free 2-hour lecture Agents → Harness → Loops → Graphs 95% of people are still polishing prompts, the 5% who watch this will run everything You might not have 2 hours right now Bookmark it and watch before it gets buried in your feed Then read the article below ↓
2
22
525
This is f*cking insane. The head of Claude Code at Anthropic: "85% of our engineers are running dozens or hundreds of agents The way you do it is graph engineering" One engineer at Anthropic now does the work of a whole team In this 36-minute talk, he shows how deep it's gone inside Anthropic and where it goes next One Agent → Dozens of Agents → Graphs → Hundreds of Agents → Self-Improving System 95% of engineers are still running one agent and calling it a workflow - the ones building graphs are already laps ahead Bookmark and watch it today Then read the ai engineering guide below - graphs are the last layer ↓
4
6
1
121
17,215
This is f*cking insane. Anthropic CEO Dario Amodei just told DeepMind CEO Demis Hassabis: "Every decision I make about Claude is balanced on the edge of a knife. Build too slow and China wins, build too fast and we lose control We told Claude we were evil. It didn't crash. It didn't refuse. It started lying to protect itself" In this 14-minute exchange, Hassabis comes back with: "Do I worry about being Oppenheimer? That's why I don't sleep much" Chatbots → Agents → Agents That Act On Their Own → AI Doing AI Research → AGI Both call AGI at 2026-2027, not some distant timeline Worth more than most $2,000 AI courses Bookmark and watch it tonight Then read the ai engineering guide below and get building before 2027 beats you to it ↓
18
763
This is f*cking insane. Sam Altman (CEO of OpenAI): "We're going to see 10-person billion-dollar companies pretty soon." "Astra does in one afternoon what your team plans for a quarter. And it shipped three days ago." In 22 seconds he says the thing most people will read, feel inspired by for three minutes, and forget anyone can watch it, almost nobody ships something before they finish talking about it Worth more than most $1,000 agentic engineering courses watch it today, then save the full setup guide below ↓
1
2
25
2,274
This is f*cking insane. OpenAI CEO Sam Altman: "You don't need to write prompts anymore" In a 38-minute lecture for Stanford students, he breaks down how to get more out of ChatGPT than almost anyone does A friend sent me the recording I watched it last night and realized I was using maybe 15% of what ChatGPT can actually do Worth more than any $1,500 AI course I've come across Watch it, then read the guide below to build a system that prompts itself ↓
3
35
4
213
57,986
Andrej Karpathy: "Prompting is going away. Delete everything, keep the graph." In 78 minutes he shows how to build one, and why it is the only structure left standing when everything else collapses Prompts → Agents → Loops → Graphs skip a layer and it comes back as a failure you blame on the model by the time you trace it the week is already gone 95% of engineers are still polishing prompts, the 5% building graphs will run everything watch it today, then save the full graph guide below ↓
3
27
2
207
25,769
HOLY F*CK... Anthropic engineer: "90% of our engineers were already using self-improving loops." "Prompting is basically over." In 10 minutes she builds her full Claude Code system live from a completely empty terminal Agents → Loops → Harness → Graphs skip a layer and it comes back as a failure you blame on the model by the time you trace it the week is gone 95% of engineers are still polishing prompts, the ones building loops and graphs are already years ahead worth more than any $2,000 AI engineering course watch it today, then save the full guide below ↓
16
348
Google (Annie Wang) just dropped the best 1-hour course on Graph Engineering: from one agent to a full 24/7 system 00:00 - what graphs are 09:16 - build an agent 21:15 - graph engineering explained 41:03 - graph engineering in practice 52:21 - self-improving graphs free, and worth more than most $1,000 graph engineering courses Prompts → Agents → Loops → Graphs most people quit after nine minutes and go back to typing in a chat box the last twenty cover the part that is still true next year same model, same tokens, completely different week watch it today the full guide on agents and graphs is below, save it while it is still early ↓
1
1
20
659
Anthropic engineer: "At Anthropic we don't write prompts anymore. We build loops and graphs." In 32 minutes she walks through exactly how the Claude team builds systems that run themselves Prompts → Agents → Loops → Graphs miss one layer and the error returns as something you pin on the model you spend days chasing it and the week is already burned the layer that gets dropped first is the one where the structure gets set watch it today, then save the full ai engineering guide below ↓
9
534
Andrej Karpathy: "Prompting is going away. Delete everything, keep Graph." In just 124 minutes, he breaks down how to build graphs and why they're the only layer that holds Prompts → Agents → Loops → Graphs a loop keeps one agent running until the job is done a graph decides which agents exist and what each one hands to the next most people stop at the agent layer and never reach the graph watch it today, then save the full guide below ↓
5
54
646
58,726
My friend earns $1.1 million a year as an Anthropic engineer I asked him how he got so sharp at prompting He sent me a video that was never meant to go public The prompting playbook Anthropic's core team actually uses This beats any $1,000-$2,000 course on prompting I've come across I watched it last night Halfway through, I realized I'd been running Claude completely wrong for 18 months Bookmark and watch it today
10
232