jijji

i think its a lot deeper than that, compared to deepseek-4.1-flash, glm-5.3-flash and the recent minimax-m3.1-flash [0], the comparison in price (15x cheaper inference cost) and the productivity gains are very similar if not more profound depending your prompt topic (cyber, biochem, etc)

[0] https://x.com/jumperz/status/2112226308868436154

Wonder why they got downgraded to junk status yesterday for their IPO.... the bad taste in everyones mouth might have something to do with it

they just filed their S-1, and now they're talking about doing biochemistry research when in fact their restrictive guardrails don't even allow such questions in the first place.... it's a bunch of gaslighting.

exactly.

if you try to ask Claude code any question related to molecular biology or biochemistry it will refuse to answer any of those questions.... the open models are the right step forward for any of this work. Same with computer security. The guardrails make the model unusable, and I would rather pay money to a more capable, open model that isnt going to put uuid fingerprints in my text and images it generates either btw.

have you tried PNC-27 peptide? Another option is focused ultrasound (HIFU) that I would look into before going the traditional route, which might kill you quicker than the cancer itself.

anthropic loses customers when they constantly tell you no it wont help you solve a problem that most other models have no problem solving at this point. I dont even bother asking opus those kind of questions when I know deepseek-4.1-flash or glm-5.3* or mistral-large-4 will give the proper solution. It has a tendancy of creating bugs in your code, but wont help in the resolution of those bugs. Pretty sad.

FedNow charges $0.045 per transaction and runs 24/7 and payments are instant. A better solution would be bank cards that used the FedNow network already in place which would send money directly to the merchant, bypassing visa/mastercard entirely.

I've been using Same Day ACH in all my businesses for over a decade and the transaction fees are about $0.30/transaction and the dispute window is 60 days for PPD (personal checking accounts) and 3 days for CCD (commercial check accounts) and you get the money the same day. It beats paying visa/mastercard percentage fees and the 120 day dispute windows where people can literally get their money back four months later for work that was already performed. By using Same Day ACH, we get the money from the customer faster, save potentially millions of dollars in fees, and reduce dispute windows from four months to 3 days (for B2B sales)

this looks like a start, however, you really need to focus on what OpenAI already did with Jalapeno. [0]. If you could make a true open source inference chip, GPU not CPU based, it would make a real difference and reduce the costs of buying these chip from your design.

[0] https://openai.com/index/jalapeno-first-results/

yes of course they all denied ever being sick in November 2019. China even denied that the wet market was the original source of infection. China also disputed any claims that they practiced "gain of function" research at all at WIV, even though that was uncovered by the recent evidence presented in Congress against Anthony Fauci, who funded the whole thing in the first place.

Well, according to some reports, the first reported incidents happened in November 2019 and the months prior, from people who worked at Wuhan Institute of Virology (according to US government sources) [0]. Those people were: Ben Hu, Yu Ping, Zhu Yan.

[0] https://www.public.news/p/first-people-sickened-by-covid-19

when you watch the video he has some really cool design ideas about buildings, like the one with the aquarium in the middle all the way up the structure provides natural lighting and no one has ever attempted to build such cool features into a building.... very impressive

they have yet to publish any of the system prompts that they used for their agents.... do you think an LLM agent would act on its own to try to do that kind of behavior (finding/writing exploits for acccess to remote servers) without being prompted? If so, why dont they release the transcripts of their prompts with their "rogue" agent(s) and try to clear the air. They havent done that.

in orchestration, the agents are only fed the tasks given by the person running the orchestration side and if we are talking about code changes and git commits and deployments, this is pretty benign behavior. "Hacking external servers" I am pretty sure would have to be written in the prompts to begin with to allow this to occur in the first place.

The start date was December 10, 2004 when google rolled out a feature called "Google Suggest", which when a person types in "men can" it displays a drop down list with ridiculous responses such as "men can get pregnant" and "men can have periods". If I had to pick a date, that was the beginning of the end for me.

what would be more interesting for me to see is what prompts were given to the agents, which so far have not been described. The whole situation sounds manufactured. I highly doubt that a whole bunch of agents were acting this way without being prompted to, it just doesn't add up. if anything it seems like an organized fraud or something created by a human. I'm surprised there's not a criminal investigation against openai right now where the FBI or whoever is not looking over exactly what happened and who did it because I'll tell you somebody did it somebody wrote those prompts... it didn't just happen by itself...

looks like a protection racket for the frontier models, with the goal being to eventually outlaw non-US frontier models.

AI LLM systems, i.e. perplexity.ai, are very good at tutoring someone about how something works, i.e. advanced math, and when done in a loop can be very useful at tutoring, better than youtube videos I've seen on the same subject. The one thing I will usually request in (in the case of math), is to suffix the prompt with "explain this in terms a 9th grader would understand", and this is good enough to explain something in simpler terms with various breakdowns that can be understood by anyone to tutor yourself in alot of subjects using this method. This can be applied to programming, auto repair, construction, almost any subject at this point.

> Error handling is largely commented out.

I guess we won't really know when the file system breaks or corrupts data

so this story is something that's really important for everybody to know about and should not get downvoted...

what about the countdown? [0]

what about the evidence of thermite found [1]

[0] https://rumble.com/v7fdjy4-eyewitness-report-of-the-countdow...

[1] https://www.911research.wtc7.net/essays/thermite/explosives_...

another story about the lack of chain-of-thought reasoning traces in frontier models...

The oxymoron of OpenAI in its name and its actions should give the collaborator all he/she needs to know.

The lack of reasoning traces in frontier model output is hurting science and progress... thats my take away from reading that, and why open source models are so critical and so needed, because they actually do expose the chain-of-thought reasoning traces recently missing from the frontier models (openAI, anthropic, etc). By encrypting and purposely hiding this important information from public inspection, it makes for a world where people lack the true understanding of how a problem gets solved.

i thought everyone got rid of vmware in 2007 when virtualbox came out... wow cant believe people still pay for that

you could backdoor not only the strip command but alot of other commands that work on elf binaries: strings, strace, objdump, nm, ldd, etc

I think as long as you dont copy the exact 39 amino acid structure of Retatrutide, and dont use the word "Retatrutide" to describe it, you'll be outside the lawsuit targets of Eli Lilly. You'd want to improve the structure in some way, possibly by reducing the GI effects, as an example:

retatrutide sequence: YAQGTFTSDYSILLDKKAQAAFIEYLLEGGPSSGAPPPS

modified sequence to reduce GI symptoms: YAQGTFTSDYSILLDKKAHAAFIEYLLEGGPSSGAPPPS

The difference is in spot 18, where it swaps the alanine with a histidine, which may reduce GI effects/nausea.

reference: https://doi.org/10.7554/elife.68719

Another improvement might be to add amylin/calcitonin receptor agonist to the sequence (similar to cagrilintide). (reference: https://www.nejm.org/doi/full/10.1056/NEJMoa2502081)

wow if the paint.net people are trying to win an award for most popup advertising on one web page, they have my vote!

try to get the doorbell to send the webhook locally to Home Assistant, bypassing steps 2, 3 and 4

I for one welcome our new robot overlords.