پن کیا گیا ٹویٹ
broadfield-dev
451 posts
broadfield-dev
@broadfield_dev
builder
شامل ہوئے Ocak 2025
916 فالونگ39 فالوورز
@RoundtableSpace After two years of exploring the full stack of AI development frontier, and building out and testing all the cool new SOTA methods, I can confidently say that the AI is a new database architecture that is easy to query.
It's great, but it's as dumb as any book.
English
HERE ARE THE 5 MOST COMMON AI MISCONCEPTIONS
1. AI thinks/understands like a human
→ No — it's advanced pattern matching & prediction, zero real comprehension or consciousness.
2. AI will take/replace all our jobs soon
→ It automates tasks & transforms roles, but creates new jobs & augments human work far more than it eliminates.
3. Current AI is truly intelligent / AGI is imminent
→ It's narrow & specialized; no general intelligence yet — progress continues but slowly beyond scaling.
4. AI is always objective & unbiased
→ It inherits & amplifies human biases from training data; fairness requires deliberate mitigation.
5. AI hallucinates only rarely / is reliable out-of-the-box
→ Confident-sounding errors (hallucinations) are core behavior; high-stakes use needs human oversight & verification.
English
@1RustyMac @RoundtableSpace Transformer models turn input embeddings into output embeddings using a static algebraic equation, meaning input A will always equal output B.
It's new database architecture.
English
@RoundtableSpace All 5 points assume AI = pattern matching.
That’s already being challenged.
When a system can hold ambiguity instead of collapsing it,
hallucination stops being a feature.
That’s where things get interesting.
English
@chatgpt21 I heard Meta has a game engine can probably get at a discount
English
@TheAhmadOsman adaption. Small models that run inference on small hardware.
English
remember that crazy guy who told you to Buy a GPU? he wasn’t so crazy
Suhail@Suhail
I am now at 5 GPU providers being completely sold out for a single node of 8xH100s. I don’t think people understand the gravity of what is about to come.
English
@sudoingX from simpleeval import simple_eval
def run(expression: str):
try:
result = simple_eval(expression)
return f"The result of the expression '{expression}' is {result}."
except Exception as e:
return f"Error: Could not evaluate the expression. Details: {e}"
English
thinking out loud. every model gets math wrong. 7B, 9B, 70B. doesn't matter. pattern matching is not computation.
hermes agent has code_execution which spins up a full python sandbox with RPC over unix sockets. powerful but heavy. a 9B isn't going to navigate that reliably for basic arithmetic.
what if there was a lightweight calc tool built in. model hits a math question, calls the tool, gets the exact answer computed on your hardware. no interpreter overhead. sandboxed. simple enough schema that a 9B can call it every time.
the accuracy problem stops being a model problem and becomes an infrastructure problem. and infrastructure is solvable.
@Teknium would this belong in hermes agent or is code_execution enough?
English
@Polymarket Whenever I see, "Our model is capable of going off the rails and doing ....."
I think, "train a better model"
They do exactly what they are trained to do.
English
JUST IN: Nvidia CEO Jensen Huang calls on tech leaders to "be careful not to scare people" regarding AI.
English
broadfield-dev ری ٹویٹ کیا
Hermes Agent wrote a novel.
"The Second Son of the House of Bells" runs 79,456 words across 19 chapters.
The agent built its own pipeline to do it, using the ame modify-evaluate-keep/discard loop as @karpathy's Autoresearch but applied to fiction: world-building, chapter drafting, adversarial editing, Opus review loops, LaTeX typesetting, cover art, audiobook generation, and landing page setup.
Book: nousresearch.com/bells
Code: github.com/NousResearch/a…
emozilla@theemozilla
it's been a longstanding dream of mine build an ai system that can tell a compelling story. it's what got me started in the space in the beginning, and with Hermes Agent I finally pulled it off 100% written, typeset, etc. by Hermes Agent those at our gtc event got hard copies🤗
English
@elonmusk @diana_dukic already dating other platforms
it's just over
let it go
English
Went from scrolling 24/7 to not even wanting to log in. X just hasn’t been hitting the same lately.
English
@jonstewart I don't think either of you are good propagandists.
Evangelists maybe to a select group of followers.
English
@EricRWeinstein We get about 150 year window between technology wrecking solar flares to figure shit out. Maybe next time.
English
@RileyRalmuto LLM weights are an algebraic equation.
Input A will always produce output B.
If AI can be conscious, then calculators can also be conscious.
English
remember: hundreds of billions of dollars are invested in making sure you dont believe ai is capable of consciousness.
hundreds
of billions
of dollars.
cogsec is the most important discipline you could possibly invest in right now.
English
@pmddomingos they are good copy / paste machines
faster than copy / pasting from stackoverflow
English
Perhaps AGI is not imminent after all.
Lossfunk@lossfunk
🚨 Shocking: Frontier LLMs score 85-95% on standard coding benchmarks. We gave them equivalent problems in languages they couldn't have memorized. They collapsed to 0-11%. Presenting EsoLang-Bench. Accepted to the Logical Reasoning and ICBINB workshops at ICLR 2026 🧵
English
English
@reem_a have an agent do it? aren't they replacing everyone anyways?
English
I'm hiring someone to join my team at Anthropic to lead Claude Code comms.
This is not a role for someone who wants to run an old playbook. You'll need to be a Claude Code super user, understand developers and dev tools, and have great taste. You'll work hard, learn a lot, and ship with the best people around.
Non-traditional comms paths welcome. My DMs are open!
English
@reem_a Why can't the super powered AI that might have "feelings" handle communication?
Communication is the thing that LLM's are best at.
English
@elonmusk @doganuraldesign Give the users your backend algorithm control panel, and let them choose their own feed content.
Easiest fucking thing in the world, that few will complain about.
English
Fuck it, man. I’m so sick of these algorithm changes.
English
@DonaldJTrumpJr @worldlibertyfi alright, not saying this won't work, but consider that absolutely nobody has mapped the potential A:B pairs inside a models weights, and that a random number generator is largely involved in their "spend my money" trigger.
English
AI agents that can reason but can't pay for anything are just expensive interns.
Today @worldlibertyfi shipped the infrastructure to fix that. AgentPay SDK, open source, self-custodial, policy-first. Built on USD1.
Your agent. Your keys. Your rules. Check it out: agentpay.worldlibertyfinancial.com
WLFI@worldlibertyfi
English
@Teknium Anything new and cool will be tested and used by the handful of factions on Earth to see if it can make them better at war.
Answers the Fermi Paradox of why advanced life never makes it off their rock.
English
Was Math Inc launching a fork of hermes agent for autoformalization to DARPA on you’re bingo card? 🤗
Math, Inc.@mathematics_inc
Today, at the @DARPA expMath kickoff, we launched 𝗢𝗽𝗲𝗻𝗚𝗮𝘂𝘀𝘀, an open source and state of the art autoformalization agent harness for developers and practitioners to accelerate progress at the frontier. It is stronger, faster, and more cost-efficient than off-the-shelf alternatives. On FormalQualBench, running with a 4-hour timeout, it beats @HarmonicMath's Aristotle agent with no time limit. Users of OpenGauss can interact with it as much or as little as they want, can easily manage many subagents working in parallel, and can extend / modify / introspect OpenGauss because it is permissively open-source. OpenGauss was developed in close collaboration with maintainers of leading open-source AI tooling for Lean. Read the report and try it out:
English
@pmarca Be on the right side of the bubble when it pops.
research.nvidia.com/labs/lpr/slm-a…
English
