From the series: Questions AI Researchers Are Still Exploring
Ever noticed how AI can sound brilliant… and then suddenly forget key details?
That’s not a bug, it’s one of the biggest challenges researchers are still trying to solve.
Here’s why AI memory is harder than it looks.
👇 Dive into the full piece
Ever noticed how AI can sound brilliant… and then suddenly forget key details?
That’s not a bug, it’s one of the biggest challenges researchers are still trying to solve.
Here’s why AI memory is harder than it looks.
👇 Dive into the full piece
LinkedIn
Why Is Memory Still Difficult for AI? | 10 Academy
From the series: Questions AI Researchers Are Still Exploring
Ever noticed how AI can sound brilliant… and then suddenly forget key details?
That’s not a bug, it’s one of the biggest challenges researchers are still trying to solve.
Here’s why AI memory…
Ever noticed how AI can sound brilliant… and then suddenly forget key details?
That’s not a bug, it’s one of the biggest challenges researchers are still trying to solve.
Here’s why AI memory…
👍2
Yesterday, we invited our Tenacious Recruitment Program trainees for a special session on 'Harness Engineering' , a conversation about what really makes AI systems useful beyond just the model itself.
One of the biggest takeaways is that AI is not just “the model.” The real power comes from the system around it: memory, tools, workflows, validation, and the human judgment guiding it all.
What made the session engaging was how it connected deep technical ideas to something everyone could relate to: technology is only valuable when it works reliably for people in the real world.
From context engineering and memory systems to human judgment and AI orchestration, the discussion challenged us to think beyond prompts, and toward building dependable, human-centered AI systems.
A meaningful way to close the final week of the TRP 1 program: learning not just how to use AI, but how to design systems around it thoughtfully.
🌐 https://10academy.org/
One of the biggest takeaways is that AI is not just “the model.” The real power comes from the system around it: memory, tools, workflows, validation, and the human judgment guiding it all.
What made the session engaging was how it connected deep technical ideas to something everyone could relate to: technology is only valuable when it works reliably for people in the real world.
From context engineering and memory systems to human judgment and AI orchestration, the discussion challenged us to think beyond prompts, and toward building dependable, human-centered AI systems.
A meaningful way to close the final week of the TRP 1 program: learning not just how to use AI, but how to design systems around it thoughtfully.
🌐 https://10academy.org/
🔥6
Hello, Everyone 👋 Question of The Day.
Which question matters most for understanding frontier LLMs?
Which question matters most for understanding frontier LLMs?
Anonymous Poll
17%
Are emergent abilities actually predictable?
21%
Can we fully interpret latent representations?
48%
Do LLMs reason or simulate reasoning?
21%
Is scale hiding fundamentally new behaviors?
From the series: Questions AI Researchers Are Still Exploring
Researchers kept making AI models bigger…
and somehow, they kept becoming dramatically more capable.
Suddenly they could write code, solve complex problems, reason through tasks, and display abilities nobody explicitly programmed into them.
And here’s the surprising part:
even AI researchers still debate why this works so well.
So what exactly happens when AI scales?
And why does “bigger” suddenly lead to behaviors that feel almost intelligent?
👇 Explore one of the biggest unanswered questions in modern AI
Researchers kept making AI models bigger…
and somehow, they kept becoming dramatically more capable.
Suddenly they could write code, solve complex problems, reason through tasks, and display abilities nobody explicitly programmed into them.
And here’s the surprising part:
even AI researchers still debate why this works so well.
So what exactly happens when AI scales?
And why does “bigger” suddenly lead to behaviors that feel almost intelligent?
👇 Explore one of the biggest unanswered questions in modern AI
❤3
Our co-founder, Yabebal Fantaye, is speaking at the AABS 2026 Connect Conference in Addis Ababa this Friday, May 15, from 12:30 PM – 01:15 PM on “Beyond Access: Preparing African Graduates for a Digital and Global Economy.”
Yabebal will be sharing how we’re thinking differently about graduate readiness,
moving beyond just access to education, and focusing on real digital fluency, global competitiveness, and innovation skills that matter for today’s economy.
You can attend his session online Via Whova app (the conference is hybrid)
For details on the full program and registration, visit https://www.aabschools.com/events/aabs-2026-connect-conference.
And if you’d like to learn more about what we’re building at 10 Academy, please do reach out and connect with Yabebal directly – he’d love to chat.
Yabebal will be sharing how we’re thinking differently about graduate readiness,
moving beyond just access to education, and focusing on real digital fluency, global competitiveness, and innovation skills that matter for today’s economy.
You can attend his session online Via Whova app (the conference is hybrid)
For details on the full program and registration, visit https://www.aabschools.com/events/aabs-2026-connect-conference.
And if you’d like to learn more about what we’re building at 10 Academy, please do reach out and connect with Yabebal directly – he’d love to chat.
❤6
Getting graduates ready for a digital, connected world isn't just about giving them access to technology, It's about working together, trying new things, and having honest, sometimes tough, conversations.
That's why we were excited to be part of the Association of African Business Schools Conference in Addis Ababa. It gave us a real chance to talk about how African schools can help students become not just digitally skilled, but truly ready to compete and innovate on a global stage.
We would like to thank Philasande Sokhela, and Cosme Almeida for sharing their thoughtful perspectives along side our Co-Founder Yabebal Fantaye on working across borders, boosting employability, and what the future of learning in Africa really looks like.
We're also grateful to the Association of African Business Schools for creating the kind of conversations that can actually shape what's next for African talent and innovation.
That's why we were excited to be part of the Association of African Business Schools Conference in Addis Ababa. It gave us a real chance to talk about how African schools can help students become not just digitally skilled, but truly ready to compete and innovate on a global stage.
We would like to thank Philasande Sokhela, and Cosme Almeida for sharing their thoughtful perspectives along side our Co-Founder Yabebal Fantaye on working across borders, boosting employability, and what the future of learning in Africa really looks like.
We're also grateful to the Association of African Business Schools for creating the kind of conversations that can actually shape what's next for African talent and innovation.
❤8👍1
🎓 Celebrating the Graduation of 10 Academy’s Tenacious Recruitment Program (TRP 1) Trainees ✨
Last Friday, we proudly celebrated the graduation of our TRP 1 trainees after months of intensive learning, collaboration, AI experimentation, late-night debugging, and continuous growth.
Throughout the journey, trainees developed not only technical skills, but also critical thinking, teamwork, and real-world problem-solving abilities.
We were honored to hear inspiring remarks from our co-founders Arun Sharma and Yabebal Fantaye, alongside industry leaders Damon Reeve, CEO of Ozone, and Deepa Manickavasagam, CEO of RewardOps, who shared insights on the future of AI, engineering, and global opportunities.
To all TRP graduates: congratulations on this incredible milestone. Your journey is only beginning, and we can’t wait to see what you build next. ✨
🎥 Watch the graduation video on YouTube:
https://www.youtube.com/watch?v=60_MoCh7smE&t=5047s
Last Friday, we proudly celebrated the graduation of our TRP 1 trainees after months of intensive learning, collaboration, AI experimentation, late-night debugging, and continuous growth.
Throughout the journey, trainees developed not only technical skills, but also critical thinking, teamwork, and real-world problem-solving abilities.
We were honored to hear inspiring remarks from our co-founders Arun Sharma and Yabebal Fantaye, alongside industry leaders Damon Reeve, CEO of Ozone, and Deepa Manickavasagam, CEO of RewardOps, who shared insights on the future of AI, engineering, and global opportunities.
To all TRP graduates: congratulations on this incredible milestone. Your journey is only beginning, and we can’t wait to see what you build next. ✨
🎥 Watch the graduation video on YouTube:
https://www.youtube.com/watch?v=60_MoCh7smE&t=5047s
❤6🔥2
Technology keeps evolving, but so do the deeper questions we ask as humans.
For this AI Meetup, we’re opening up a thoughtful conversation around AI and spirituality, how technology intersects with meaning, belief, consciousness, and the human experience.
Come for an open dialogue, shared perspectives, and a welcoming space for curiosity.
🗓 May 20, 2026 (Wednesday)
⏰ 6:30 PM EAT
📍Around 22, Comet Building (next to Axum Hotel), 2nd Floor, Office 205, Addis Ababa
Come join the conversation and feel free to bring a friend!
For this AI Meetup, we’re opening up a thoughtful conversation around AI and spirituality, how technology intersects with meaning, belief, consciousness, and the human experience.
Come for an open dialogue, shared perspectives, and a welcoming space for curiosity.
🗓 May 20, 2026 (Wednesday)
⏰ 6:30 PM EAT
📍Around 22, Comet Building (next to Axum Hotel), 2nd Floor, Office 205, Addis Ababa
Come join the conversation and feel free to bring a friend!
❤7
Hello builders 👀 Question of the day.
You inherit an ML system with incredible accuracy, low latency, and great offline evaluation. Six months later, the business impact is still disappointing. What’s the MOST likely hidden issue?
You inherit an ML system with incredible accuracy, low latency, and great offline evaluation. Six months later, the business impact is still disappointing. What’s the MOST likely hidden issue?
Anonymous Poll
38%
Wrong optimization objective
17%
Poor feature/data semantics
26%
Feedback loop distortions in production
29%
Model decisions aren’t interpretable enough
Hello AI engineers 🙌 Champions League scenario!
You’re building a real-time analytics system for a Champions League match ingesting:
- Live event streams (passes, shots, fouls, xG)
- Multi-camera tracking data (25 frames/sec per player)
- Social sentiment feeds (X + TikTok reactions)
- Betting odds updates (every few seconds)
During a high-stakes game:
👉 Ingestion latency is stable
👉 But different regions show different “live match momentum” at the same timestamp
Stack:
Kafka + Spark Streaming + Feature Store + Cached APIs
You’re building a real-time analytics system for a Champions League match ingesting:
- Live event streams (passes, shots, fouls, xG)
- Multi-camera tracking data (25 frames/sec per player)
- Social sentiment feeds (X + TikTok reactions)
- Betting odds updates (every few seconds)
During a high-stakes game:
👉 Ingestion latency is stable
👉 But different regions show different “live match momentum” at the same timestamp
Stack:
Kafka + Spark Streaming + Feature Store + Cached APIs
🔥7❤1