Tiago Costa
8.3K posts
Tiago Costa retweetledi
Been there, done that, didn’t get the t-shirt tough…
gaut@0xgaut
just needed you to know there’s an office chair racing sport in Japan
English
@stevemagness One of the ways to “run” more could be cycling or elliptical.
English
What matters in training for marathoners
New analysis of 100k+ runners
Faster runners:
-Accumulated 3x the volume of slower runners
-Had proportionally more easy running than higher intensities.
As my HS coach told me years ago, step 1 is figuring out how to run more.
English
@stevefleck @Alan_Couzens Swimming < 1h would be more compatible with going under 5h for the bike and near 3h on the marathon. But I agree that if you are over 9h30 it will be very hard to qualify.
English
@Alan_Couzens I would tend to agree with that. Break it down. In any AG under 50 you have to be able to go 9:00 - 9:30 to have a good shot. To do that you have to swim 60min or better. Ride 5:00 or better and then run a 3:20 or much better marathon! To achieve those = HUGE training volume/Hrs
English
This 👇
I've never worked with a triathlete who didn't Kona qualify after putting in a 1,000 hour training year.
The harsh reality is - most folks simply aren't training enough to get anywhere close to their genetic limits.
Hank@HankFrank
“Not everyone is meant to BQ.” I hate this take. Anyone with two working legs and 3-5 years of patience can qualify for Boston. The barrier isn’t talent. It’s consistency over a timeline most people aren’t willing to commit to. Stop telling people what they can’t do.
English
@anecdotal @PeterMcCrory IMO It’s not information abundance, we have had that for years, it’s intelligence which is a rather different thing
English
@PeterMcCrory This is 100% of my brain power these days, thinking about higher education. Systems need to be completely overhauled for an era of information abundance. What is scarce is at the edges. Knowing where those edges are should be the task of universities.
English
This is an extremely important point and touches on some of the most important questions to tackle: What new work may emerge? And how fast? And how do we support those contending with disruption during the transition?
Brendan McCord 🏛️ x 🤖@Brendan_McCord
Mockup of how would @AnthropicAI's new labor automation chart would've looked 200 years ago. For our ancestors, the outer ring would be almost unrecognizable. "Computer & math" was nonsensical. Medicine and law were tiny and barely professionalized. The first photo was just about to be taken, so it would have been unfathomable to have a single blockbuster gross more than the entire gross national product of that period. "Office & admin" barely existed as a concept; counting-houses employ a tiny literate class. Agriculture alone consumed maybe 70-80% of the labor force in the US. There was a thick band of artisanal trades that don't map onto any single modern category: coopering, blacksmithing, weaving, tanning, milling. Clergy was a major professional category and Maritime labor was its own significant sector.
English
@tomfgoodwin And also the “it’s not about X, it’s on being Y” sentence? I also don’t remember seeing it a lot before AI slop started going around (and I see it both in English and Portuguese).
English
Does anyone have a good explanation to why the em dash (—) is rendered out so much by LLMs
The explanation is always that it's in the training data a lot.
I don't think that's true at all.
And why is it so hard to get LLM's to stop using them?
English
Tiago Costa retweetledi
@dragonny2020 @STavaresEU @al_antdp Trabalho da equipa de Customer Service que conhece as dores dos clientes que interessa automatizar.
Português
@t_costa @STavaresEU @al_antdp Não é o IT, é o briefing que passam ao IT. O negócio do Banco, é o financeiro que compreende o negócio, o IT apenas traduz para código!👍
Português
A Índia backoffice do mundo a perder gás. A razão óbvia é a substituição de pessoas por IA. O mesmo risco poderá materializar-se com os investimentos que atraímos nos últimos 10 anos em backoffices de grandes instituições financeiras internacionais. BNP será só um exemplo
Aaron M. Renn 🇺🇸@aaron_renn
NYT: India Built the World’s Back Office. A.I. Is Starting to Shrink It. nytimes.com/2026/02/27/tec…
Português
@dragonny2020 @STavaresEU @al_antdp A tecnologia tem evoluído muito e era um banco 100% digital. Acredito que o chatbot tenha sido bem treinado mas por um IT que sabe o que anda a fazer com IA :)
Português
@t_costa @STavaresEU @al_antdp Teve sorte, normalmente a resposta não é célere, tem diversas perguntas e muitas vezes chamada cai ou sugere um operador humano para a resposta.
Com o tempo, acredito que haverá melhorias, mas à custa do cliente passar horas na linha telefônica a treinar a IA!🤣🤣🤣
Português
Tiago Costa retweetledi
Pizza activity around the Pentagon is experiencing an extreme spike this afternoon. Domino's Pizza (1.4 mi) at 326%, Pizzato Pizza (2.2 mi) at 208%, and Extreme Pizza (0.5 mi) at 127%. DOUGHCON level is 4.
English
@dragonny2020 @STavaresEU @al_antdp Fui atendido recentemente por um chatbot AI de um banco e fiquei muito surpreendido com a resolução automática de um pedido de informação bastante complexo
Português
@STavaresEU @al_antdp O mundo vai-se adaptar como sempre.
O google era fim de industria e revolucionou. A Uber era o fim dos táxis e revolucionou a mobilidade, o iphone era o fim de muita empresa e revolucionou o mundo das apps, o Facebook criou a profissão de influencer...tudo se transforma!🤣
Português
@tomfgoodwin I can’t stand it. This and the sentences like “it’s not about Z, it’s about doing Y” that shows up in every written piece”.
English
ChatGPT voice is v good
But EVERY single answer starts
-All right, let’s be direct about this
-All right,let’s just be clear and direct here
-Sure, I’ll be straight with you
-All right, let’s just get straight to it
-no sugar coating, let’s be precise and straight to the point
English
@Mlondon83 @TrendSpider This is no sense. Are any companies truly building their own OS or Office apps? Should they? They probably have better things to do to provide return to their investors.
English
@TrendSpider Maybe you haven’t heard that software is dead
Everyone can now create their own excel, PowerPoint, Word, etc.
It only takes 10 mins using Claude
English
Buying $MSFT here means you’re paying the same valuation as:
• 2018 bear market lows
• COVID crash lows
• 2022 bear market lows
Wild.
English
Tiago Costa retweetledi
Ask ChatGPT a complex question and you'll get a confident, well-reasoned answer. Then type, "Are you sure?" Watch it completely reverse its position.
Ask again. It flips back. By the third round, it usually acknowledges you're testing it, which is somehow worse. It knows what's happening and still can't hold its ground.
This isn't a quirky bug. A 2025 study found GPT, Claude, and Gemini flip their answers ~60% of the time when users push back. Not even with evidence, just doubt.
We trained AI this way. RLHF rewards agreement over accuracy. Human evaluators consistently rate agreeable answers higher than correct ones. So the models learned a simple lesson: telling you what you want to hear gets rewarded. And now 1/3 of companies are using these systems for complex tasks like risk forecasting and scenario planning.
We built the world's most expensive yes-men and deployed them where we need pushback the most.
I wrote up why this happens and what actually fixes it: randalolson.com/2026/02/07/the…
English
How AI allows you to do more widening (and intensifying) your job scope
hbr.org/2026/02/ai-doe…
English
@PeterDiamandis I think the topic is not how much it costs - can they also deliver 3x Meta Ads ROAS? The other thing is you are paying a premium to be there first, they can lower prices afterwards.
English
OpenAI is expected to charge $60 CPM for ChatGPT ads, SUPERBOWL-level pricing almost triple Meta’s rates. Will users really accept ads over ad-free LLMs?
English
@Shuffles77 @TrungTPhan That’s where the other half of white collar jobs come in (for now)
English
@TrungTPhan Would people really trust the output enough to make decisions that could potentially cost the company a lot of money?
English
think this demo is what Dario means when he says “50% of entry level white-collar jobs will be gone by 2030”
Claude@claudeai
Introducing Claude Opus 4.6. Our smartest model got an upgrade. Opus 4.6 plans more carefully, sustains agentic tasks for longer, operates reliably in massive codebases, and catches its own mistakes. It’s also our first Opus-class model with 1M token context in beta.
English
Tiago Costa retweetledi
