Margaret (@Mrsalways_) header
Margaret (@Mrsalways_) profile picture

30’s , millennial , Aries♈ building my little digital empire soft life, big dreams🎀

New Jersey
Joined October 2023
Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)Photo by Margaret (@Mrsalways_)
businesses don't need the most powerful ai model for every task sometimes the smarter investment is choosing one that gets the job done without running up the bill with google's gemini 4 carbon entering the market, is the added capability worth the added cost?
6
having your finances turned into interactive charts sounds useful now if chatgpt could explain why my money disappears faster than i earn it, we'd really be getting somewhere
OpenAI expanded its ChatGPT Finance feature to be available on ChatGPT desktop. With ChatGPT Finance, users can also generate interactive widgets based on their financial data. * Finances is available to Free, Go, Plus, and Pro users in the U.S.
5
Margaret retweeted
SemiAnalysis now spends 45% of its AI budget on persistent agents: CEO Dylan Patel said the firm mostly runs Perplexity Computer, calling Anthropic's and OpenAI's own Slack agents a letdown.
19
10
136
31,516
businesses are using more ai, but their spending isn't necessarily keeping pace getting more value without constantly increasing the budget sounds like a win. especially when companies are under pressure to show a return on every dollar spent
2
six months without a paycheck can wipe out years of savings and the longer people stay unemployed, the harder it becomes to keep up with everyday expenses
🚨Job hunting is getting BRUTAL in America Nearly 2 million people have now been jobless for over 6 months, up 85% since early 2023. The average job search now takes 27.1 weeks, up 9.3% in just one month. In the last 78 years, long term unemployment has only been this high during or after a recession.
3
Margaret retweeted
BREAKING: US intelligence says Nvidia CEO Jensen Huang had a "romantic relationship" with a Chinese spy 20% chance he's out as $NVDA CEO before 2029
Readers added context they thought people might want to know
US intelligence has issued no public confirmation; the allegation is based on media reports citing anonymous sources about a classified memo. Nvidia has described interactions as routine business. afr.com/technology/us-… nytimes.com/2026/11/06/us/… engadget.com/2310084/china-…
275
207
223
2,787
1,986,538
headline says an ai agent wiped a company’s whole database in 9 seconds. fastest migration in history, wrong direction. give the agent dev credentials only. it can’t drop what it can’t reach.
9
github killed premium requests june 1. copilot runs on ai credits now, one credit is a cent. pro is still $10, pro+ $39. the flat-fee agent buffet is over. check your usage page before the invoice does it for you.
5
omg y’all github copilot has /rewind 😭 it restores the conversation AND the files copilot changed, even without git. action: type /rewind when it goes sideways. con: it restores files copilot changed, so keep your own edits committed
17
y'all claude projects are free now. up to 5. upload your docs + instructions once, and every chat keeps the context. claude.ai → projects → new project → add files + instructions 🍵
10
Margaret retweeted
legora an AI startup focused on legal work is approaching a $10 billion valuation. most people outside fintech and professional services have not heard of legora. that is why it is interesting. legal work is one of the clearest AI use cases that nobody outside the industry talks about. the documents are long, structured, and repetitive. the review process is expensive and time consuming. the stakes for accuracy are high. and the billable hour model means every minute saved is directly measurable in dollars. legora builds AI that reads contracts, flags risks, identifies inconsistencies, and summarises legal documents. it does not replace lawyers it reduces the time spent on the low-value parts of legal work so lawyers can focus on the high-value judgement calls. the $10 billion valuation from a company most people have never heard of reflects a simple observation: when you automate work that costs $500-800 an hour at law firms, the ROI case writes itself. this is what professional services AI looks like when it works. not a chatbot. not a writing assistant. a system that does the expensive, tedious parts of a specific workflow so the expert can focus on the parts that require expertise. the investor thesis is that every major law firm and in house legal team becomes a customer. at those price points, a small number of customers is a large amount of revenue.
4
8
107
omg stop re-explaining your job to chatgpt every chat 😭 settings > personalization > custom instructions, two boxes, save. tip: don't mix contradicting tones, they fight each other. set once, done forever. screenshot: your own settings page shows exactly where.
13
used chatgpt's codex for a backend fix last week. it didn't just autocomplete, it actually reasoned through the bug. still checked every line myself. tools don't replace judgment, they just clear the noise around it.
13
Margaret retweeted
We've added Kimi K3 to Perplexity and Perplexity Computer for Pro and Max subscribers. Kimi K3 in Perplexity is hosted exclusively on U.S.-based servers.
74
65
29
1,011
135,478
AI divas hallucinating again, hunty! , Fake citations, bogus books, invented casesfact-checked & WRONG. Ask 3 ways: inconsistent flops. Verify everything! (112 chars)
14
Just hit 10M weekly users. Codex + ChatGPT Work went from 2M to 10M in months.We’ve entered the Agentic Era. Developers aren’t chatting — they’re shipping.Speed + DX = rocket fuel. Who’s building with AI daily? 👇
11
AI: Machines doing human-like tasks.LLMs (like Grok): Language AI trained on trillions of tokens via Transformers. Predict next token for chat, code & summaries.Strengths: Creative. Limits: Hallucinations, no real understanding.
8
big 2026 shift: ditch complex multi-agent setups.a single strong LLM agent with deep discussion → PRD → /goal mode often beats them all. less complexity. way more output. future of productivity = better alignment, fewer agents. who simplified and saw big gains?
5
2026 game-changer: stop just chatting with AI. build real workflows with agents & memory. one system now handles my inbox, content, research, no more context switching chaos. the leap isn’t better prompts. It’s systems. building your first agentic flow this month? 👇
1
9
after 194 days of tweaking, I finally emptied my claude. deleted folders, about-me files, global instructions, and every template. now it’s just Skills + projects. less noise = way sharper outputs. minimalist AI setup just hits different who else runs a clean claude?
6