Emmanuel Acheampong

AMI and Yann LeCun are betting the real future of AI is specialized. Their take: world models built on JEPA that reason causally, plan under constraints, and work in messy real-world environments. Their Feb 2026 paper coins it: Superhuman Adaptable Intelligence (SAI). The idea in short is “Outperform humans on specific tasks, tackle domains we simply can’t.” Nabla (clinical AI) is already on board as AMI’s first partner. More architectural innovation instead of more compute. Curious to see where this goes. https://lnkd.in/gmB4k4EA

19

2 Comments

Co-founders Sid Pardeshi and Brian Elliott explore Anthropic's "circuit tracing" - a technique that reveals the specific computational pathways and decision-making processes happening inside AI models, giving us unprecedented visibility into the "why" behind AI responses. Full conversation on YT @Blitzyai #circuittracing #anthropic

7

1 Comment

What if your AI workloads ran faster and at a lower cost? At #AdvancingAI Tareq Amin, CEO of HUMAIN, shares a bold promise: a 30% lower TCO no matter what your current best is. Challenge accepted! (and Lisa Su’s in on it too).

119

Shout out to Sanmi Koyejo! He's a professor at Stanford University and a president of Black in AI, Koyejo focuses on "Trustworthy Machine Learning." He works on the structural integrity of AI models - ensuring they work reliably across different populations and demographics. "My lab’s algorithms help ensure that AI systems diagnosing diseases from chest X-rays work equally well for patients of all racial backgrounds, preventing health care disparities" (quoted in Stanford news, 2025; also tied to broader fairness in medical imaging papers he co-authors or influences). He shifts emphasis from overall accuracy to subgroup fairness (group performance parity), addressing real issues like underdiagnosis in underserved groups (e.g., racial minorities in public datasets like CheXpert/MIMIC-CXR). His work on algorithmic fairness and domain adaptation is the blueprint many companies reference as they build out AI solutions. #BHM #ai #tech #responsibleai

2

🧠💬 Memory in LLMs A practical guide showing how to implement conversational memory in LLMs using LangGraph, demonstrated through a therapy chatbot. Features code examples for basic retention, trimming, and summarization approaches. Learn to build memory-aware apps 👉 https://lnkd.in/gybcrV5v

967

21 Comments

San Francisco startup mixus is betting on a hybrid AI model where humans remain strategic collaborators, not just fail-safes. Co-founded by Elliot Katz and Shai Magzimof, Mixus builds “colleague-in-the-loop” AI systems that escalate only high-risk decisions—such as financial approvals or policy enforcement—to human oversight, keeping automation efficient while ensuring accountability. Used by enterprises like major retailers for tasks like anomaly detection, Mixus integrates seamlessly with Slack, Google Drive, and enterprise tools via the Model Context Protocol (MCP). The platform allows customizable oversight rules and transparent workflows. This approach responds to growing concerns over AI reliability—Salesforce reports AI agents succeed in only 35% of complex tasks—and joins a broader industry trend toward human-AI collaboration, also embraced by OpenAI and Slack. Though critics warn that oversight can introduce inefficiencies and automation bias, Mixus sees human supervision as a strategic advantage. The company forecasts exponential growth in AI agent deployment by 2030, with human overseers playing a pivotal role in managing scale, trust, and compliance.

2

What a builds strong AI engineers is sequence, not shortcuts. Foundations first. Patterns next. Models later. Deployment last. Miss a layer and everything feels harder than it should. We care about clarity in how AI is built, learned, and scaled across Africa and the diaspora. Because real progress comes from structure. Wants to become a AI/ML engineer this year? Follow this sequence. #ArtificialIntelligence #MachineLearning #AIEngineering

3

2 Comments

Good breakdown by Jason re: OpenAI's recent write up of their big data & context infrastructure they've been working on. Why this is important for enterprises: 1. If OpenAI needed to develop a purpose-built data context management system to improve internal use of data, then it's probably not good enough for you to just connect data to a foundational model (or chatbot) and expect results. 2. Wiring data context together is complex, nuanced, and requires serious engineering, context engineering, and several disciplines across AI / ML / behavioral science to get right (or you could work with Puller AI ). 3. Updating your semantic tools manually is also out of the question. Let's start from the assumption that schemas drift, data stays disparate, your catalogs and semantic layers will always be imperfect, and expertise will need to somehow be ingested from the brains and behaviors of SMEs. That requires constant, AI-scale maintenance. 4. Buyer beware. Every AI agent and app in 2026 will claim to have their own data context graph -- begs the question: Will enterprises want context to be fragmented and managed at the edges? (Rhetorical, the answer is no). Favorite quote from the blogpost: Meaning Lives in Code. As Data Context Management becomes a core capability for AI-first organizations, we'll see increased reliance on AI to stitch together and maintain data context (as opposed to manual upkeep). That's why Puller AI automatically reviews, ingests, and organizes the world of implicit context that can be inferred in your code & workflows. DM me if you'd like to chat about how to unlock your company's Data Context Management capabilities.

7

Interview Kickstart has launched an Advanced Generative AI course to help engineers and data professionals build hands-on expertise with LLMs, diffusion models, and multimodal AI systems. Designed for working professionals, the program focuses on real-world model design, deployment, and capstone projects that reflect how generative AI is being used across modern engineering teams. Explore the full news here: https://lnkd.in/ddcYDkup #GenerativeAI #InterviewKickstart #AIUpskilling #MachineLearning #LLMs #DiffusionModels #TechCareers #FutureOfWork #hrtechpulse #bestplacetowork

8

Allen Zhu (GSR Ventures): The AI app boom is imminent — the next ByteDance, Kuaishou, or Rednote was likely founded this year. At the 2025 Inclusion Bund Conference, Allen Zhu delivered a clear, unsentimental read on where AI is headed: the real opportunities are beyond the model. Key takeaways for builders & investors Model plateau → miniaturization. Under the Transformer paradigm, core “intelligence” gains are tapering; wins will come from UX and cost. Expect smaller, cheaper models, smarter data curation, and edge deployment. Apps are shifting text → voice → video. Token consumption is exploding in real use cases. Voice is breaking out now; ultra-low-latency video (~1s) could trigger the next wave. Moats live outside AI. Think workflows, editing/polish, proprietary/regulated data, operations, and specialized hardware (e.g., capture devices, smart name badges). Do the “boring, dirty work” others avoid. Agents: go vertical or get absorbed. Generic agents lack defensibility and will be eaten by advancing base models. Build sticky, domain-specific agents tied to real processes and outcomes. Commercialization > vibes. Retention is king. Beware “vibe revenue.” In SV, many VCs wait for ~$2M ARR post-launch; in China, sub-12-month paths to ~$5M ARR are becoming the bar. China founders’ edge & go-global. Strength in To-C experience design and hardware supply chains; look to Japan/SEA/MENA to avoid direct “model wars” and compete on execution. From UNIQUE RESEARCH’s tracking of thousands of AI apps globally, we see the same arc: the model dividend is flattening; winners are those who embed AI into specific scenarios, compress latency/cost, and build non-technical moats in data, workflow, and user insight. Bottom line: Don’t chase generic model horsepower. Ship products that deliver outcomes in real workflows—and measure yourself by retention, renewal, and unit economics. https://lnkd.in/gC63AABv

2