Jason Marcieski (@JasonMarcieski) header
Jason Marcieski (@JasonMarcieski) profile picture

Apeironist

Pittsburgh, PA
Joined May 2014
Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)Photo by Jason Marcieski (@JasonMarcieski)
Jason Marcieski retweeted
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 these systems" In 30 minutes, she walks through exactly how Anthropic builds agents that remember, fix their own mistakes and get sharper with every run. Better than any $1000-$2000 agent course I've seen. Watch it, then read the article below ↓
Readers added context they thought people might want to know
Lamis never says "You're not supposed to prompt Opus 5.5” in here talk at AI Dev Con. The caption misrepresents the video knowing most won’t watch the video all the way through and jump to the articles. youtube.com/watch?v=tTcxVv…
218
666
43
8,816
1,649,695
Jason Marcieski retweeted
Call me crazy (cuckoo cuckoo)… but I will keep repeating this Claude + instagram is going to make more millionaires in 2026 than crypto ever did in 2015
624
386
5
3,060
296,364
Jason Marcieski retweeted
this is pure f*cking treasure A Stanford AI research group found how to orchestrate Claude Opus 5.5 and GPT-6.1 Sol together, and it completely breaks the search scaffolding wall most developers try to combine frontier models by chaining prompts in one context window or hardcoding rigid evolutionary loops. you burn tokens on context bloat and freeze models into fixed search rules that out-date themselves within 5 iterations Stanford's architecture eliminates human-designed search scaffolding and splits the cognitive stack: > search director: Claude Opus 5.5 plans the search, querying a persistent idea-graph where MAP-Elites and MCTS reduce to single Cypher queries > candidate proposer: GPT-6.1 Sol explores high-entropy code variations inside isolated execution sandboxes > zero context bleeding: sessions reset after each iteration to prevent prompt bloat while the graph stores lineage > evaluator isolation: candidate code never touches the scoring process, stopping reward hacking and test leakage cold > meta-agent distillation: an offline meta-pass extracts verified lessons and updates search guidance across branches the benchmark metrics outclass standard discovery baselines: > 3.2x lower model spend than fixed evolutionary frameworks > 1st place rank across 7 competitive AtCoder heuristic contests against human competitors > beats published SOTA on Anthropic kernel builder and 11 mathematical optimization tasks > 0% reward hacks: candidate-controlled metric tampering eliminated Opus directs the search graph. Sol explores the code you stop writing rigid search harnesses. you let frontier models own the discovery loop
137
597
34
3,467
282,693
Jason Marcieski retweeted
Anthropic opened up a ribosome study and found the machine stalls on purpose, with the pauses doing a job nobody assigned them. Speed was always the metric. Faster translation, better output, and a stall counted as something gone wrong. > TIME - translation tracked second by second instead of measured end to end > MARK - every pause logged with the exact point in the sequence it happened > SORT - random hesitations separated from pauses that repeat across cells > FOLD - each repeating pause matched against what the protein was doing then The pauses are not noise. The same ones land at the same places, in cell after cell, far more often than chance allows. They cluster where the protein needs a moment. A section finishes folding before the next one is fed out behind it. Rush that and the chain comes out complete and useless, folded into a shape that does the wrong thing. So the stall is the instruction. The sequence carries timing information on top of the amino acids everyone reads it for. Optimising genes for speed has been standard practice for decades. It quietly deletes that timing. That explains a lot of engineered proteins that are technically correct and biologically dead. Every repeating pause and what it was waiting for sit in the breakdown below.
56
241
11
1,674
266,705
Jason Marcieski retweeted
Someone literally put an entire AI engineering degree on GitHub and made it 100% free 🔥 523 lessons across 20 phases, with ~342 hours of material. The whole curriculum is built around learning AI from first principles. You start with the math, then implement the concepts yourself: backpropagation, tokenizers, attention, agent loops and more. Production frameworks come later, once you've already built the core mechanics by hand. Every lesson runs these 5 steps: Read the problem → derive the math → write the code → run the test → keep the artifact. By the end of each lesson you have reusable: prompts, SKILLmd files, agent definitions, MCP servers you built yourself. The code runs in Python, TypeScript, Rust or Julia, depending on what fits the concept best. The curriculum goes from linear algebra all the way to autonomous agent swarms, multi-agent coordination, production infrastructure, and AI safety. It’s 100% free, open source, MIT licensed and available to clone and run locally. 65k+ stars on GitHub. aiengineeringfromscratch.com
2
292
7
1,821
111,299
Jason Marcieski retweeted
You can basically work with agents to write markdown skills and put it on a cron job and do almost every useful type of knowledge work now
Alexandr Wang, Meta's chief AI officer: "Internally at Meta, we have seen cases where, if you develop the right agentic loop and have the right evaluation system and metric for the agents to optimize, a swarm of agents can accomplish more than a team of 100 engineers. They can do it very handily, actually, very easily." Wang was in conversation with Garry Tan, head of Y Combinator, at Startup School 2026. Interested in AI? Follow @AiEvolutio58513 and never fall behind. I track ChatGPT, Claude, and every tool quietly changing how we work and create, then hand you the tested signals.
134
177
24
3,127
443,776
Jason Marcieski retweeted
A billionaire sat in a room for 42 minutes and listed every psychological trick that makes people lose money. For free. The finance industry has spent thirty years pretending this recording doesn’t exist. He didn’t sell a course. He didn’t write a newsletter. He sat in a chair at 96 years old and explained why brilliant people do the dumbest things with their money. Then he explained why they’ll keep doing it. MBA programs charge $200,000 to teach behavioral finance. He covered 25 biases in one sitting. Some of them still aren’t in any curriculum. He gave the entire framework away on camera. The part nobody talks about: he called crypto antisocial. He said index funds will crush most managers. He said private equity is full of wretched excess. He said all of this in a room full of people who manage money for a living. Nobody argued. A hedge fund analyst at a top firm told me this is the first thing they send to anyone who joins the desk. Not a book. Not a model. A 42-minute video of a 96-year-old man explaining why you’ll be wrong and how to recognize it before it costs you everything. 40 million people have heard his name. Almost none of them have watched him explain the 25 ways their own brains are working against them. The lecture is free. He died the following year. It’s in the video.
16
276
4
949
135,207
Jason Marcieski retweeted
if you have too many interests, this 35 min Dan Koe video will reshape what you do with your curiosity: >you have 10 interests >everyone tells you “pick one” >but specialists get replaced first by AI >so you keep all 10 anyway >do nothing with them >you’re not lazy >you just have no plan >plans bring out clarity out of anxiety >your 10 interests connect >turns you into a unique combo >that’s not easily replaceable
11
143
1
1,340
63,347
Jason Marcieski retweeted
It worked. Gave Claude $20. Left it coding overnight. Woke up to: • finished script • +$95 from first 5 trades Didn’t touch anything. Then I found this wallet: 0xb27b...b82 • $384,858 profit • 13,070 predictions • $18.5K biggest win • Started March 2026 Almost no one is watching it. While building my script, I copytraded it to test the math. Started with $80. → $4,218 in a few weeks Every position is identical in structure. Not trading. One equation. Running while you sleep. Examples: $17,839 → $36,318 (+103%) $3,112 → $15,011 (+382%) $10,676 → $22,178 (+107%) Here’s the exact framework I gave Claude: 1. Temporal bias Find BTC time windows with >65% edge. Ignore everything else. 2. Kelly sizing Size based on confidence. Never fixed bets. f* = (p × b − q) / b 3. Volatility filter Skip high volatility. Trade only clean conditions. 4. Bayesian updates Update probabilities after each trade. The system improves over time. Claude built everything while I slept. Funded it with $20. First trade fired minutes later. You’re seeing this early. That window doesn’t stay open. You only need Claude + Device + 1 hour/day. Giving This Free for 24 hours. To get it: 1. Comment the word ' DM ' 2. Like and Retweet this post 3. Follow me @TechElyra
273
169
4
356
41,981
Jason Marcieski retweeted
This guy literally explained why some people become successful… while others stay average. And the reason is uncomfortable. It’s not just hard work. It’s not luck. It’s Game Theory. Once you understand this, you’ll start seeing the “game” behind success differently. Watch this. 👀
99
1,015
14
4,579
1,167,147
Jason Marcieski retweeted
It is a black magic circle that traps infinity. Electrical engineers use this geometric labyrinth to tame the chaos of high-frequency signals. Long before digital solvers existed, the Smith Chart emerged as a graphical masterpiece. It takes the boundless plane of complex impedance and collapses it into a single bounded disk. Heavy arithmetic for transmission lines becomes simple visual geometry. Engineers plot resistance and reactance along these sweeping arcs. This lets them instantly navigate reflection coefficients and standing waves. Phillip H. Smith conceived this elegant trap in 1939. It remains an indispensable map for the unseen currents of our modern wireless world.
60
437
43
2,503
63,928
Jason Marcieski retweeted
Anthropic senior engineer just released a 1-hour course on building a team of agents with loops & graphs: • 00:27 - introduction to CLAUDE.md & Plan mode • 11:24 - building "skills" & "hooks" for Claude agents • 37:02 - building agents & subagents with Claude • 52:47 - self-improving loops & graphs for Claude agents this 1-hour watch will replace a $500 agentic engineering course watch today, then read how to build a team of self-improving agents that work together
54
711
13
3,975
532,533
Lockwood Archer - the start-up experiment begins. #vornova #comsic7 #blockorb
38
Jason Marcieski retweeted
Google's CEO just revealed why 2026-2030 is the last opportunity for regular people to get rich. Here’s exactly what he said… & how you can capitalize:
87
396
16
2,912
1,144,346
Jason Marcieski retweeted
Junior vs Senior
4
43
1
408
12,019