Sebastián Marambio (@SebasOnEarth) header
Sebastián Marambio (@SebasOnEarth) profile picture

Makers gonna make. I’m on a mission to make online learning more social and build technology that brings us closer together.

Joined June 2010
Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)Photo by Sebastián Marambio (@SebasOnEarth)
Dictating is much faster than typing, and transcribing meetings has become far more useful now that you can give the transcript to your agent right after the call and have it follow through on what you agreed to. Willow Voice, Tactiq, Granola, Fireflies, etc. There are a dozen companies in this space. Some do dictation, others do meeting transcripts. Personally I've been paying for both Willow and Tactiq. Willow is $28 per user per month on the team plan and Tactiq is $29. For my team of 6 that's $4,104 a year, for tokens that would cost maybe ~$200 if you paid OpenAI directly, probably less. If one of these companies had solved both dictation and meeting transcripts I would have kept paying for the convenience. But you know what's even more convenient? Vibe coding your dream app in a couple of hours. So here it is. Free and opensource for macOS. You bring your own tokens. And let me know if you want it for Windows or Linux and I'll tell my agent to help you with that lol I built it with Sol 6.1, and Opus 5.5 made this video on the first go. What an amazing time this is. github.com/brainforwarding/t…
6
6
275
I had early access to this game and have been playing it for a few months. You turn over cards, fight monsters, find treasures and get stronger as you go. You only have three moves, so it looks simple, but choosing which one to use next can be surprisingly hard. A game takes about 20 minutes. There’s a daily challenge where everyone gets the same game and only one try. Then you can see who made it out alive and compare scores 🔥
1
1
123
the future will be amazing AI doomers just happen to be getting more attention right now
1
1
73
apparently all it takes is a two-month internship at Anthropic to become a prophet of the apocalypse
37
I've been using Grok Bots for the last couple of weeks. I have not tried it for coding but i'm using it to keep track of X and curate content for me. Two things that stand our are: 1. the useful setup is specialization plus a chief-of-staff bot that prompts the other agents and keeps the team moving. 2. these bots run 24/7 on a spaceX machine. Until now, keeping agents moving when the computer sleeps was hard for anyone who was not a power user. Now it's so easy!
SpaceXAI engineer (ex-Cursor): "I don’t prompt my ~20 GrokBot agents. I built a 24/7 system once - now it automates 90% of my life and prompts for me Instead of just delegating tasks like 99% of people - I delegate the delegation. One Chief of Staff gets the other bots working - result: a system managing 950K leads" In this 55-min course, SpaceXAI shows the shift of using GrokBot and her best workflow This is worth more than spending 10 hours on other courses Watch how SpaceXAI builds GrokBot - then learn all rules for such system and run it ↓
2
194
El Salvador is showing promising results from its education pilot. Of course this is getting attention in Chile, where almost everything Bukele does becomes national news. But it would be a mistake to turn this into a debate about Bukele. These results are not isolated. In Nigeria, a six-week generative AI tutoring program improved student outcomes by about 0.3 standard deviations. In Ghana, one hour a week with an AI math tutor improved scores by 0.36 SD. In the US, Tutor CoPilot increased the share of students mastering math topics by 4 percentage points, and by 9 points for students working with less experienced tutors. A 2026 World Bank review of 14 randomized studies across ten economies found that adaptive and AI-enabled education programs improved learning by an average of 0.125 SD. These are big impacts! 0.3 SD as in the case of Nigeria or gana is roughly 1 to 2 years of typical learning progress. But the evidence also shows that AI can hurt learning when it is poorly designed. One experiment found that students with unrestricted access to GPT performed better while using it, but 17% worse on a later exam without it. A version designed with teacher input and learning safeguards avoided that harm. The lesson is the way AI interventions are designed and implemented matters. I expect AI and adaptive technology to become useful tools for teachers who want to provide more personalized instruction. They can help teachers give each student practice, feedback, and support at the right level. Chile is in a good position to test these ideas. TALIS 2024 reports that 55% of Chilean teachers already use AI in their work, compared with 36% across OECD education systems. Chilean schools also have relatively strong access to technology and connectivity. We should not adopt one program across the whole country based on results from somewhere else. But we should not spend another year trying to design the perfect experiment before doing anything either. Start with schools that want to participate. Decide what outcomes matter. Measure them carefully. Learn what works and what does not. Improve the program as we go. Then run larger studies and scale the approaches that continue to produce good results. AI is changing quickly. Our way of testing it needs to let us learn quickly too.
A few days ago, I shared the World Bank’s summary of the results of the PISA for Schools assessment conducted in 171 Salvadoran public schools participating in our education pilot. Since then, I have seen some legitimate questions about what these results mean, as well as some claims about what they supposedly don’t mean. So let’s be very precise. First, what we NEVER claimed: 1. We never claimed these results represent the national average. They don’t. This was not the national PISA assessment. It was PISA for Schools, conducted specifically in the 171 public schools participating in our pilot program. That is precisely why my original post said: “These are the results of our pilot program, which includes AI tutors and has been running in 171 public schools for just over a year.” A pilot is, by definition, not the entire country. 2. We never claimed that these 171 schools were a statistically representative sample of every school in El Salvador. They weren’t selected randomly from all Salvadoran schools. They were selected to participate in the pilot because they were schools where the program could initially be implemented more easily: they had certain basic infrastructure and connectivity requirements, and school communities capable and willing to implement a new educational model. That creates a selection effect, and it should be acknowledged. But there is an important distinction here. These schools were NOT selected because their students had exceptional academic results. They were not our “best schools.” Not even close. We didn’t search the country for the highest-performing schools or students and then test them. They are normal Salvadoran public schools, urban and rural, in different parts of the country, serving ordinary Salvadoran families. And the characteristics that made implementation easier: connectivity, infrastructure, trained teachers and willingness to adopt the program, are also things that can be replicated in other schools. They are not innate characteristics of the students. We can install connectivity. We can improve infrastructure. We can train teachers. We can provide the same technology. We can deploy the same educational tools. In fact, that is exactly what we are doing. 3. These results also do NOT prove that AI tutors, by themselves, caused the improvement, although we believe they are a MAJOR part of these results. Our pilot is broader than that. It includes changes in teaching, support for teachers and principals, educational materials, monitoring and evaluation, technology, and AI tools, among other interventions. Our AI tutors and learning platforms are being developed with @grok, @Kira_Learning, and support from I Hate Flying Bugs. And the broader transformation of our education system is being supported by the @WorldBankGroup, @AgendaCAF, and @el_BID. All of these interventions can also affect student motivation and engagement. But that distinction matters: if students become more motivated BECAUSE of the program, that is not necessarily a flaw in the program or an explanation that somehow invalidates the results. It may be one of the ways the program works. The same applies to teachers. For example, the implementation of the AI tutors, including the requirement for teachers to log in and out of the system, coincided with teacher attendance in these schools increasing from 76% to 97%. If the program causes teachers to attend school more consistently, and that in turn helps students learn more, that is not an advantage that disappears when we scale the program. It is potentially part of the intervention itself. Of course, to scientifically isolate the causal effect of AI tutors, teacher attendance, student motivation, technology, training, or any other individual component would require a different experimental design. Now, what DO these results tell us? 1. They tell us that 1,198 students in 171 ordinary Salvadoran public schools… 🧵1/2👇🏼
8
73
1
329
34,412
Here are the sources I mention Nigeria From Chalkboards to Chatbots World Bank blogs.worldbank.org/en/educa… Ghana Effective and Scalable Math Support AI math tutoring through WhatsApp arxiv.org/pdf/2402.09809 US Tutor CoPilot AI support for tutors during live instruction edworkingpapers.com/sites/de… India Disrupting Education? Personalized technology-aided instruction aeaweb.org/articles?id=10.12… Cross-country review Can EdTech Close Learning Gaps? World Bank review of 14 randomized studies thedocs.worldbank.org/en/doc… Turkey Generative AI Without Guardrails Can Harm Learning pnas.org/doi/10.1073/pnas.24… And the figure on AI use by Chilean teachers comes from TALIS 2024 oecd.org/en/publications/res…
1
2
289
"Everyone is declining" is the wrong reading The headline everywhere today is that PISA scores fell to their lowest level ever recorded. That is true of the OECD average, which is a different claim from saying that every system declined. Of the 68 systems with a comparable decade-long trend in science, 27 rose, 39 fell and 2 stayed flat. Forty per cent went up during the same decade in which the average went down. The spread deserves more attention than the midpoint. Cambodia gained 72 points per decade and Türkiye 62, in a year when the OECD average moved by −7. Spending does not sort the two groups either. Luxembourg spends USD 288,521 per student across a school career and scores below Ireland, France and Italy, all of which spend less than half that amount. The OECD describes PISA as something that "is not a league table, it provides a compass." Using it that way means asking where the systems that improved actually got to, and how they did it, instead of tracking the average position of everyone standing still.
53
added sync to kanban.page now you can scan a qr with your phone and the board is on both devices no account or login, nothing to sign up for. end-to-end encrypted, so the relay can't read a word of it
1
41
My Salesforce arc: → Felt sorry for anyone who had to learn to use it → Claude came out, was sure Salesforce was dead → Salesforce integrated Claude, now I want to try it → Matthew McConaughey made the ad, now I’m sharing it What a comeback 😂
Welcome Claudeforce. ❤️
1
1
396
some guy just dropped a kanban app that works better than Trello and it's totally free, no sign up. you can just start using it, and everything stores privately in your own browser it's open source so you can fork it, add more features and make it your own *some guy = me + Claude, mostly Claude
2
2
2,673
for over a decade i was a hardcore fan of google docs i have not used it for over a year now that ai is here, markdown files in your own computer is the way to go there are two alternatives: - obsidian-like app + an agent (cowork, antigravity) - vs code (integrates both)
9
2
1
62
9,193
Are you looking for a part-time teaching gig? Here's what we are up to in case you are interested 👇
1
129
👩🏻‍🏫 Interested? We're creating a scalable model where more TAs can join. It's practical, it's growing, and it could reshape how online learning works.
1
95
Right now we need TAs for courses on: - AI for education - Evaluation - Susteinability - Digital skills - Educational innovation
83