rhahn.eth

1.5K posts

rhahn.eth banner
rhahn.eth

rhahn.eth

@RogerHahn

🌐 CEO @KLIK | 🎓Patent Lawyer, Engineer, Finance | 🚀 Tech Entrepreneur | 🎙 Global Speaker | 🔥 BioE, BME, MBA, JD

Manhattan, NY Beigetreten Kasım 2015
812 Folgt906 Follower
Angehefteter Tweet
rhahn.eth
rhahn.eth@RogerHahn·
MEMENTO AI: When Your AI Has Amnesia 🧠💾 What if I told you... your AI has the same problem as Leonard from Memento? 🎬 "Remember Sammy Jenkins... Remember Your Context Window"
rhahn.eth tweet media
English
2
0
0
2.4K
rhahn.eth
rhahn.eth@RogerHahn·
@MarioNawfal Why is this guy shouting the whole time. Relax and make your point calmly.
English
0
0
0
6
Mario Nawfal
Mario Nawfal@MarioNawfal·
Trump: Iran Ceasefire Is“ON MASSIVE LIFE SUPPORT” - w/ Fmr Intelligence Officer Stefano Ritondale x.com/i/broadcasts/1…
English
145
49
161
370.8K
divyansh tiwari
divyansh tiwari@DivyanshT91162·
GitHub may have just killed vibe coding. Their new repo “spec-kit” already has 92k+ stars — and it reveals where AI-driven development is actually heading. Instead of telling your AI: “Build me a todo app” and hoping for the best… You run 6 commands that turn your idea into an executable specification: • "/speckit.constitution" → defines project rules (quality, testing, UX) • "/speckit.specify" → explains WHAT to build, not the tech • "/speckit.clarify" → AI asks questions to remove ambiguity • "/speckit.plan" → choose the stack and architecture • "/speckit.tasks" → generates dependency-ordered tasks • "/speckit.implement" → the agent builds it The deliverable is no longer the code. It’s a living specification your AI can read, debate, and execute. Works with Claude Code, Copilot, Cursor, Codex, Gemini, and 25+ other agents. The shift most people still don’t see: “AI writes code” → “AI executes specifications.” Intent-driven development is the next era of software development. Repo👇
English
69
160
1.5K
162.4K
rhahn.eth
rhahn.eth@RogerHahn·
@heyrobinai You don’t understand how Le world model works. It sits on top of generative models 🙄
English
0
0
0
18
Robin Delta
Robin Delta@heyrobinai·
THE ENTIRE AI INDUSTRY JUST GOT HUMILIATED a tiny model trained in just a few hours on a single graphics card is planning 48x faster than billion-dollar supercomputers. It actually understands physics instead of just memorizing patterns. yann lecun was right the whole time for three years every major lab told you the same story. scale is all you need. just throw more GPUs at it. just train on more tokens. eventually the model will "wake up" and understand the world. it was a lie. or at minimum, a very expensive bet that just lost. LeCun kept saying generative AI is a dead end. predicting the next pixel or the next token is fundamentally wasteful, the model burns trillions of parameters memorizing surface details instead of learning how reality actually works. he proposed JEPA instead. predict abstract concepts in a compressed thought space. don't paint the world pixel by pixel, understand it. the problem was JEPA kept collapsing. left to its own devices the model would cheat, mapping a dog, a car, and a human to the same point in latent space. technically minimizes the loss. learns absolutely nothing. every fix was ugly. seven loss terms. frozen encoders. EMA tricks. stop-gradients. the kind of duct-tape engineering that should have been a red flag. then LeCun's team dropped LeWorldModel. they replaced all the hacks with one regularizer that forces the latent space into a gaussian distribution. the model can no longer cheat. to make accurate predictions it has to actually encode physics. 15 million parameters. single GPU. trains in hours. plans 48x faster than foundation world models. detects physically impossible events on its own. meanwhile OpenAI is raising another $40B to train GPT-6 on a data center the size of manhattan. the entire scaling thesis just got embarrassed by a model that fits on a gaming PC.
Robin Delta tweet media
English
219
697
3K
255.4K
Mario Nawfal
Mario Nawfal@MarioNawfal·
🇮🇳 damn, the guy tosses those cars around like it's nothing tragically one person died during the attack. Another was seriously wounded
English
52
47
435
52.1K
rhahn.eth
rhahn.eth@RogerHahn·
@milesdeutscher Haiku for plan mode? should do other way around. Max on plan (use write-enhanced-spec) with red teaming, then use sonnet for well scoped tasks, connected to the architecture I mentioned above.
English
0
0
1
56
rhahn.eth
rhahn.eth@RogerHahn·
@milesdeutscher Too simplistic. I recommend Hermes, wiki, a single source of truth (SsoT) and custom slash commands with MCP “brain” running on local docker. Create a context intelligence layer for each slash command using a self hosted PostGres db that stores your prompt history on docker.
English
2
0
3
1.8K
rhahn.eth
rhahn.eth@RogerHahn·
@sukh_saroy Facts. Recursive self improvement feels like a scam unless the labs are hiding some incredible models. I feel the Claude writes 80% of the code narrative is marketing fluff because there is a 100% human effort to keep alignment.
English
0
0
0
207
rhahn.eth
rhahn.eth@RogerHahn·
@lifeof_jer Railway is the problem. Opus 4.6 did the same thing to me on a test project but never anywhere else. But your story doesn’t pass the sniff test. Who puts production data on Railway much less for 5 years. Getting strong vibe coding. Even your post is blatantly AI.
English
0
0
0
123
rhahn.eth
rhahn.eth@RogerHahn·
@MarioNawfal Well trained. Thinking of cross fire. He didn’t want to shoot his own guys.
English
0
0
0
1.5K
rhahn.eth
rhahn.eth@RogerHahn·
@MarioNawfal Are you in the news business or the tarot card business?
English
0
0
0
11
Mario Nawfal
Mario Nawfal@MarioNawfal·
I've been saying the war is about to end since April 2 and the war is over since the ceasefire, but I have a new caveat: IF IRAN MAINTAINS MAXIMALIST DEMANDS, U.S. OR ISRAEL MAY ASSASSINATE MORE OF THEIR LEADERS Although this is highly unlikely Let's see how negotiations go
Mario Nawfal@MarioNawfal

Someone is trying to community note my post below saying I declared 'the war is over 6 times in the last month' I take that as a compliment 😂 Here's the full breakdown of my prediction posts: April 2 (before Trump's address to the Nation): "The war is close to ending" (At that time everyone was expecting the U.S. to invade Iran, calling it 'imminent'. I saw it as a bluff to pressure Iran to end the war, as an invasion made no sense to me, and all military experts I spoke to agreed) April 2 (after Trump’s address to the Nation): "I REPEAT, THE WAR IS CLOSE TO ENDING" (The markets crashed and oil spiked after the speech as everyone thought he was escalating when he said he will bomb Iran "back to the Stone Ages", but I saw it as the clear beginning of the off-ramp) April 5: "Trump’s threats are a bluff" (This is when Trump began threatening to bomb Iran's infrastructure, including their power plants and bridges, and posted "Open the Fin' Strait, you crazy b***s, or you'll be living in Hell — JUST WATCH!". Again, I saw it as a very clear off-ramp, as these threats constituted a war crime, and I did not think Trump was insane) April 6: "Both Trump and Iran want to end the war, the only question is - on who’s terms" (Trump continued his posts, reaffirming my belief the war was aboutn to end) April 7: "off-ramp incoming" ; "A good leader knows when to walk away" ; "Trump will not nuke Iran" (Hours before the deadline, Trump posted "Open the F***in’ Strait... or you’ll be living in Hell" and "a whole civilization will die tonight". While people like Tucker freaked out pleading with Trump's advisors to take away the nuclear codes, I saw this as the perfect execution of the "Madman Theory") April 8: THE WAR IS CLOSE TO ENDING (I made this statement as the deadline was approaching. Hours later we were the first to break the news globally that a ceasefire was signed) April 9: Expect Lebanon to get a ceasefire (This is when the U.S. and Iran were at odds on whether Lebanon was included in the ceasefire. I knew Iran would get their way, they won the war and would therefore not abandon their proxy Hezbollah as they did in 2024-25) April 9: Lebanon war ends that week (As Netanyahu said the war against Hezbollah will continue and Lebanon is not part of the ceasefire, I knew he would eventually have to capitulate to Trump's demands. I've always said: Israel has too much influence over the U.S., but they do not have the final say, especially not with Trump) April 9: Ceasefire will hold, war is over (Many expected the Iran ceasefire to break down because of Israeli actions and rhetoric in Lebanon) April 9: Lebanon war is over (right after Israel bombed Lebanon very hard) (Many expected Israel's very heavy bombardment of Lebanon that day was an attempt to break the ceasefire, while I saw it as Netanyahu's last ditch attempt to cause Hezbollah as much damage as possible before having to abide by Trump's ceasefire) April 14: War is over (I made it clear I expect the ceasefire to hold. I repeated this statement yesterday, as you can see below, as the U.S. fired at and boarded an Iranian ship) WHY I THINK THE WAR IS OVER: The reason I've made all these optimistic posts is simple: I know Trump didn't want a long war, I know he doesn't want the global economy to collapse, I'm also well aware of the limited munition stockpiles the U.S. has, and I also don't think Iran want to keep seeing their country bombarded. It was clear to me all along that this war, one I thought would never happen, was one big stupid miscalculation. I also didn't think Trump just went insane as some people feared. That was too lazy and illogical of an explanation. The "Madman Theory" made a lot more sense to me considering his actions and words throughout this war. Now does that mean I am certain the war is over? Absolutely not. There's still a high risk that both sides don't reach a deal, Iran is emboldened by their successes this past month, and Trump does not want to get embarrassed, so there are many ways this can go wrong. Iran's IRGC need to accept some conscessions to give Trump a POLITICAL win. Trump has already accepted Iran won the war, and that they will come out in a stronger regional position at the detriment of Israel, and I think he's fine with that, as long as it's not embarrassing for him politically. I do however believe more pragmatic heads will prevail, and at most we will only see limited strikes before a permanent peace deal. I included all the links to my posts above in the comments below, and you can also see them in the HIGHLIGHTS section of my account, I keep them all. Hope this helps

English
57
3
95
48.6K
How To AI
How To AI@HowToAI_·
Yann LeCun was right the entire time. And generative AI might be a dead end. For the last three years, the entire industry has been obsessed with building bigger LLMs. Trillions of parameters. Billions in compute. The theory was simple: if you make the model big enough, it will eventually understand how the world works. Yann LeCun said that was stupid. He argued that generative AI is fundamentally inefficient. When an AI predicts the next word, or generates the next pixel, it wastes massive amounts of compute on surface-level details. It memorizes patterns instead of learning the actual physics of reality. He proposed a different path: JEPA (Joint-Embedding Predictive Architecture). Instead of forcing the AI to paint the world pixel by pixel, JEPA forces it to predict abstract concepts. It predicts what happens next in a compressed "thought space." But for years, JEPA had a fatal flaw. It suffered from "representation collapse." Because the AI was allowed to simplify reality, it would cheat. It would simplify everything so much that a dog, a car, and a human all looked identical. It learned nothing. To fix it, engineers had to use insanely complex hacks, frozen encoders, and massive compute overheads. Until today. Researchers just dropped a paper called "LeWorldModel" (LeWM). They completely solved the collapse problem. They replaced the complex engineering hacks with a single, elegant mathematical regularizer. It forces the AI's internal "thoughts" into a perfect Gaussian distribution. The AI can no longer cheat. It is forced to understand the physical structure of reality to make its predictions. The results completely rewrite the economics of AI. LeWM didn't need a massive, centralized supercomputer. It has just 15 million parameters. It trains on a single, standard GPU in a few hours. Yet it plans 48x faster than massive foundation world models. It intrinsically understands physics. It instantly detects impossible events. We spent billions trying to force massive server farms to memorize the internet. Now, a tiny model running locally on a single graphics card is actually learning how the real world works.
How To AI tweet media
English
431
2.1K
12.2K
1.3M
GREG ISENBERG
GREG ISENBERG@gregisenberg·
There’s $1T up for grabs for agent-first startups and this window is WIDE open. Probably 10,000+ niches. How it plays out: 1. Every SaaS company follows salesforce and goes headless within 18 months 2. a new category of "agent-native" startups emerges that treat salesforce, HubSpot, workday etc as dumb backends. the startup IS the agent. the SaaS is just the database. 3. the entire consulting/services industry around enterprise SaaS gets compressed into software. the agent replaces the implementation team. 4. outcome-based pricing becomes default. nobody pays per seat when the "seat" is an agent making 10,000 API calls a minute. you pay when revenue hits your account. 5. the winning founders are ex-operators who understand a vertical workflow cold. the code is the easy part. knowing that a property manager spends 14 hours a week on lease renewals? that's the insight worth $100M. 6. distribution becomes the moat. when anyone can wire agents to APIs, the company with the audience and the brand wins. media + agents is the new SaaS. There’s a rush to incubate live/short form shows. 7. Silicon Valley goes all influencer. Roy lee gets this. Pat Walls gets this. Sam Parr gets this. 8. the first $1B agent-native company in each vertical will look nothing like the SaaS it replaced. smaller team, higher margins, no implementation cost, no churn from bad UX because there is no UX. the fastest path to wealth right now: find an industry that still runs on dashboards, phone calls, and spreadsheets. build the agent-native version. charge per outcome. own the workflow end-to-end. someone reading this right now is going to build a $100M company off this exact shift. tell me about it on the @startupideaspod when you do. Im rooting for you. Less reading, less bookmarking, more building. the last wave rewarded people who built pretty interfaces on top of ugly data. I think this wave rewards people who build smart agents on top of exposed APIs. Or who just build the APIs themselves Here we go
Marc Benioff@Benioff

Welcome Salesforce Headless 360: No Browser Required! Our API is the UI. Entire Salesforce & Agentforce & Slack platforms are now exposed as APIs, MCP, & CLI. All AI agents can access data, workflows, and tasks directly in Slack, Voice, or anywhere else with Salesforce Headless 360. Faster builds, agentic everything. 🚀 #Salesforce #Agentforce #AI venturebeat.com/ai/salesforce-…

English
186
268
3.1K
579.1K
rhahn.eth
rhahn.eth@RogerHahn·
Claude is getting just stupid. First, weekly rate limits. Then, 4 hour window rate limits at peak hours. And now, RANDOM rate limits for normal workloads! @theo is right! At this point, Claude is basically unusable! Anthropic didn't invest in compute Open AI be gloating
rhahn.eth tweet media
English
0
0
0
65
rhahn.eth
rhahn.eth@RogerHahn·
@MarioNawfal Amazing how AI slop in 2026 is way better than a Disney movie
English
0
0
0
160
Mario Nawfal
Mario Nawfal@MarioNawfal·
Cargo ships on their way through the Strait of Hormuz after both Iran and the U.S. blockaded it
English
30
27
448
58.9K
rhahn.eth
rhahn.eth@RogerHahn·
@JohnnyNel_ My recs - MCP server to postgres "brain" running on Docker local with a "heartbeat" /loop. All context aware. Beats Openclaw any day of the week /code-debugger /browser-brain /factory-system /refactor LangGraph implemation github.com/rhahn28/the-fo…
English
0
0
0
15
Johnny Nel | AI for Founders
@RogerHahn rate limits are brutal when you're building... context windows matter way more than most founders realize for automation
English
1
0
0
31
rhahn.eth
rhahn.eth@RogerHahn·
Piling on Claude usage limiting rant. Pain in the🍑and breaks flow state. Been with Claude before it was cool - MCP, shitty models, skills.md, Desktop, Code, every feat. release. But for first time in long time, I'm testing alternatives. #anthropic #claude
English
1
0
1
63
Clarence
Clarence@choto_clarence·
#AbuDhabi authorities are responding to fires at the Borouge petrochemicals plant, one of the world's largest integrated polyolefin complexes, after debris from intercepted missiles and drones caused damage. Operations are suspended, and emergency teams are on the scene #Iran #US
Clarence tweet media
Clarence@choto_clarence

The US Immigration and Customs Enforcement (ICE) arrested Hamideh Soleimani Afshar, niece of Qassem Soleimani, and her daughter, after Secretary of State Marco Rubio revoked their green cards. Afshar, a vocal supporter of the #Iranian regime, had lived in the #US for years #Trump

English
1
1
2
536
rhahn.eth
rhahn.eth@RogerHahn·
@GoogleAI Sam and Dario just shat their pants
English
0
0
0
42
Google AI
Google AI@GoogleAI·
Today, we’re launching Gemma 4, our most intelligent open models to date. Built with the same breakthrough technology as Gemini 3, Gemma 4 brings advanced reasoning to your personal hardware and devices. Here’s what Gemma 4 unlocks for developers: — Intelligence-per-parameter: Our 31B (Dense) and 26B (MoE) models deliver state-of-the-art performance for their size, outcompeting models 20x their size on @arena — Commercial flexibility: Released under a permissive Apache 2.0 license for complete developer flexibility and digital sovereignty — Agentic workflows: Native support for function-calling and structured JSON output allows you to build reliable, autonomous agents — Multimodal edge AI: The E2B and E4B models bring native vision, audio, and low latency to mobile and IoT devices — Long-context reasoning: Up to 256K context windows allow you to process entire repositories or large documents in a single prompt Whether you're building global applications in 140+ languages or local-first AI code assistants, Gemma 4 is built to be your foundation. Explore in @GoogleAIStudio or download the weights on @HuggingFace, @Kaggle, and @Ollama.
English
149
355
2.4K
1.7M