𝐍𝐢𝐟𝐞𝐦𝐢 𝐀𝐥𝐩𝐢𝐧𝐞

This One ChatGPT Prompt Lets AI Apply to Every LinkedIn Job for You! GPT PROMPT: “Apply to all visible jobs on this linkedin page using the Easy Apply option. Open each job, click Easy Apply, and follow the specific instructions for that application. Do not rely on a fixed sequence, as each job may have different steps. If prompted to select a resume, choose the appropriate one. If asked whether you’re authorized to work outside your country, select “No.” If asked if you require a visa, select “Yes.” At the final review step, scroll within the modal to reveal and click the “Submit application” button. Never skip or discard any application. Make sure each is successfully submitted before proceeding. If prompted to save and exit, click the “X” icon to resume. Scroll the job list on the left to load more jobs when needed. Continue applying without stopping until all jobs are completed.”

@ryan_tech_lab @codr_x i doubt any post training was done its simple just a very expansive and improved system prompt may be they tweaked via post training based on learnings from system prompt improvements so its easy to replicate

The frustration is valid. But it is worth separating the RL fine-tuning from the base model. Cursor post-trained specifically for agentic code tasks, and that does change the behavior. The real issue is that they marketed it as something proprietary when it was built on public infrastructure. That is the trust problem.

i use composer heavily on cursor and its quite disappointing that i am paying to use a free model (kimi k2) time to resurrect @codr_x

companies like openai, anthropic, and a few other model providers intermittently degrade their model outputs they intentionally lower the output quality so users are pushed to retry prompts, burn more tokens, and drive more usage and profits all in the name of corporate greed

founders: “i need funding to get traction." investors: "you need traction to get funding."

that's his clone btw

large language models can’t create new knowledge. they’re trained on existing data and generate outputs by predicting patterns within that data. claims that this technology will independently produce groundbreaking inventions or scientific discoveries go beyond what it currently does. the system recombines information it has already seen. it does not originate ideas outside that scope. despite this, ai labs and leaders like sam altman, elon musk, and others continue to suggest that ai will drive major breakthroughs in science and medicine, which overstates the present capabilities of llm-based systems. however, because llms can efficiently reproduce and recombine what already exists, they are well suited to automating many current knowledge-based tasks. much of modern work is built on processing information, following patterns, and producing variations of existing formats. that means the immediate and more realistic impact of these systems is not revolutionary invention, but replacement. if deployed at scale, these models could displace large portions of the workforce in writing, coding, design, customer support, analysis, and other fields where output is pattern-based. the likely short-term result is cost reduction for companies and increased returns for capital owners. the longer-term consequence could be widespread job loss and downward pressure on incomes if new forms of work do not emerge at the same pace.

isn't this just reintroducing quadratic costs in depth? for ultra-deep models, block compression might bottleneck nuanced retrieval why do this?

at some point vibe coders will learn that just because you can ship something doesn’t mean you should

just a glorified gpt wrapper pretending to be a large language model

this is insane 🥳

see results 🔗 browsergpt.co/share/extract-…

i was able to automate extracting dozens of tweet captions and urls from endless scrolling pages using browsergpt by civai @aibrowsergpt. result is instant data collection saving hours of manual work if done manually see results

@HoopsCrave 😂😂😂

Kanye West on wishing Kim the best: "I can't wish her the best because she would end up with me"

people at the helm of ai are like babies in adult bodies. they can swing between extremes, most times extremely lucky and at other times surprisingly naive

"i’m running twenty openclaw subagents in parallel, each with their own customized models, contexts and specialized tasks" the agents:

you can order food and other items from chowdeck directly through browsergpt by civai on desktop and mobile coming soon to @aibrowsergpt apps 🚀🍔📦

attention has and will always be end game

@AIBrowserGPT 🔗 browsergpt.co/share/use-quic…

i was able to automate selecting thousands of checkboxes across multiple pages using browsergpt by civai (@aibrowsergpt). result is instant bulk actions saving hours of manual work if done manually, see results in replies