DigDeep Tech Podcast

About DigDeep Tech Podcast

The latest in gadgets, gaming and tech! digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast)

Publisher

Ajay Cyril

Category

business · technology

Language

en

Explicit

No

First episode

1 Jun 2024

Latest episode

8 Mar 2026

Latest episodes

9 episodes in the feed.

1. 

   8 Mar 2026

   What Agentic AI Actually Is

   Everyone is talking about agentic AI. You hear it in conference talks, product announcements, and investor decks. Every major tech company seems to be launching some version of “AI agents.” But if you ask ten people what agentic AI actually means, you will probably get ten different answers. Most explanations fall into two categories. Either they are extremely technical and hard to follow, or they are vague marketing language that never really explains how these systems work. So instead of starting with definitions, let’s start with a simple scenario. A Simple Example: Processing Refunds Imagine a company that processes 10,000 refund requests every month. There are three main ways to run that operation. The first is humans. A support agent reads the message, finds the order in the CRM, checks the refund policy, and triggers the refund. It works well, but it scales linearly. If refund volume doubles, you need to hire more people. At typical support salaries, this works out to roughly $4 to $6 per refund. The second approach is RPA. Robotic process automation tools follow predefined workflows. A bot opens the inbox, extracts an order ID, checks a rule, triggers the refund, and sends a confirmation email. This works well when everything is structured. But real life rarely stays structured. Customers send screenshots instead of order numbers. Policies change. APIs evolve. Data formats drift. When that happens, the automation breaks. RPA handles structure. Reality introduces variability. This is where agentic systems enter the picture. What Agentic Systems Actually Do Instead of following a rigid script, an agentic system receives a goal. For example: Resolve this refund request. From there, the system reasons through the workflow. It interprets the customer message, searches the CRM for relevant orders, retrieves the latest refund policy, evaluates eligibility, and then calls the appropriate API to issue the refund. If the refund exceeds a certain threshold, it routes the decision for approval. The key difference is subtle but important. RPA executes predefined steps. Agentic systems reason through goals. The Real Insight: Architecture But the most important thing to understand is this. Agentic AI is not just a model. The model is only one layer of the system. At the center sits a probabilistic reasoning engine. When given a task, the model predicts the next action, then the next, and then the next. This creates a loop that looks something like this: Plan.Execute.Evaluate.Adjust. That loop is what makes agentic systems feel intelligent. But enterprises cannot run critical systems purely on probability. They require guarantees. That tension between probabilistic reasoning and deterministic infrastructure defines the architecture of agentic AI. The Architecture Behind Agentic AI In production systems, the model is wrapped inside several layers of infrastructure. First is retrieval. Instead of relying on the model’s training data, the system retrieves live information from enterprise systems like CRMs, databases, and knowledge bases. This grounds the model in real company data. Second is tool execution. The model does not directly manipulate systems. Instead, it proposes structured actions. These actions are validated before they are executed through APIs. Third is identity. Agents inherit the permissions of the user who triggered them. If a support agent cannot approve a $10,000 refund, the agent cannot either. Fourth is policy enforcement. Business rules live outside the model. Even if the model predicts approval, the system can block the action if it violates policy. Finally, there are execution traces. Every decision made by the system is recorded. This allows engineers to audit, replay, and analyze how the system behaved. Together, these layers transform a probabilistic model into a system that can operate inside real enterprise infrastructure. Why This Matters The shift toward agentic systems is not just a technical change. It is also an economic one. A human-driven refund workflow might cost around five dollars per request. RPA might reduce that to around three dollars. Agentic systems can push that closer to two dollars. When companies process hundreds of thousands or millions of transactions, that difference becomes significant. But the bigger change is flexibility. RPA automates steps. Agentic systems automate reasoning within constraints. That makes them far more resilient in environments where data, policies, and workflows constantly evolve. A Better Mental Model The simplest way to think about agentic AI is this: It is a probabilistic reasoning engine operating inside deterministic guardrails. Without those guardrails, you have a demo. With them, you have production infrastructure. And that is why so many companies are suddenly rebuilding workflows around AI agents. Watch the Full Breakdown I recently made a short video breaking down this architecture visually. Final Thought RPA automated steps. Agentic systems automate bounded reasoning. And when the cost of reasoning drops, the architecture of work starts to change. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

2. 

   13 Dec 2025

   Five takeaways from AWS re:Invent 2025

   Most re:Invent recaps focus on product launches. That misses the real story. AWS used re:Invent 2025 to show how second movers win long term. Graviton changed the cost structure Graviton is no longer an alternative CPU. It is the default. When the cloud provider itself runs cheaper infrastructure, pricing power compounds quietly over time. Trainium is about economics, not trophies AWS is not chasing peak benchmark numbers. It is creating predictable training costs while keeping GPU compatibility. Enterprises care about budgets, not scorecards. Nova models prioritize reliability Workflow completion matters more than chat quality. Nova models are optimized for consistency, not cleverness. Agents are built for operations Most agent systems fail when governance appears. AWS built memory, policy, and auditability first. AI Factories expand where AI can run Sovereign launchpads show how AI can move into regulated environments without fragmenting the stack. AWS was not first in AI hype. It was patient. That patience is now turning into control. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

3. 

   6 Dec 2025

   Stop Buying the Wrong Mesh WiFi: The Only Guide You Need

   The Mesh WiFi Trap: What UAE Homes Are Getting Wrong Most UAE homes are upgrading their WiFi the wrong way. The top mesh WiFi search results look ideal, but each of them has a hidden limitation. Here’s a clear, useful guide to choosing the right system for your home. Why the Top Mesh WiFi Results Are Misleading When you search for “mesh WiFi” on Amazon or any major e-commerce platform, the first few results look perfect. Attractive pricing, clean branding and strong claims like “WiFi 6” or “WiFi 7.” But each of those products has a catch. One system advertises WiFi 7 but performs like an entry-level WiFi 6 router. Another is priced at a level only large multi-storey villas actually need. And one is labelled WiFi 6 even though the underlying technology is still WiFi 5. These traps are incredibly common, and most people end up overpaying or choosing a system that does not solve the real issue. Let’s break down what you should actually look for. The Only Naming Rule You Need: AC vs AX vs BE This is the simplest and most important decoder when buying WiFi equipment: AC = WiFi 5 AX = WiFi 6 BE = WiFi 7 If the model name starts with AC, it is older technology. Many listings still label AC devices as “WiFi 6” for SEO optimisation, which creates confusion. Once you internalise this naming rule, 70 percent of bad purchases disappear. Dual-Band vs Tri-Band: Why It Matters in UAE Homes The UAE has a very specific challenge that many other countries do not: concrete walls. Concrete causes significant signal loss. A mesh system that performs well in an American wood-frame home may collapse entirely in a Dubai or Abu Dhabi villa. This is where the band configuration matters. Dual-band systems are adequate for apartments. Tri-band systems are much better for villas. The reason is simple: tri-band gives the mesh nodes a dedicated backhaul channel. That channel prevents interference and ensures that each node gets a clean, stable connection. The difference in performance is dramatic in multi-room homes. The Setup That Actually Improves Your WiFi Most people buy a mesh system and simply place the nodes randomly. That almost always leads to disappointment. Here’s the setup that consistently works: 1. Connect your main mesh node to your Etisalat or du router using a CAT8 cable. This single step eliminates the bottleneck that slows down most mesh networks. 2. Place the second node halfway between your router and the areas where you use the internet most. Think of it as a midpoint relay. 3. Add a third node only if there are still dead zones. Most homes don’t need more than two nodes when the placement is correct. This simple layout has a bigger impact than upgrading to a more expensive system. The Two Mesh Systems That Make Sense in 2025 After months of testing across different homes, these two stand out as the most practical and reliable for UAE users: TP-Link Deco X20 AX1800 (WiFi 6) A dependable, cost-effective choice for apartments and small-to-medium homes. WiFi 6 performance, stable mesh behaviour, and easy setup. ASUS BD4 (WiFi 7) A future-proof, high-performance option suited for villas, large homes, and users with high device counts or gigabit-plus plans. This is the right choice if you want longevity and consistent speed across every room. You don’t need to spend thousands or buy overly complex hardware. These two systems, combined with correct placement, fix almost every WiFi issue most people face. Final Thoughts WiFi problems feel mysterious, but the solutions are simple. Identify the right standard, choose the correct band type for your home structure and fix the backhaul. A few smart decisions outperform expensive hardware. If you’re planning to upgrade your WiFi, this guide should save you both money and frustration. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

4. 

   15 Nov 2025

   Apple Is Paying Google to Fix Siri. What That Really Means

   Apple is paying Google about 1B dollars per year to power the next version of Siri. That alone captures how unusual this moment is. A company built on privacy and control is now licensing intelligence from the world’s largest advertising company. Bloomberg’s report was clear. Apple Intelligence slipped behind schedule. Siri could not deliver reliable multi-step actions. Several promised iPhone 16 features never reached users. Inside Apple, the strain was obvious and top engineers left. At the same time, the market signaled something important. Most users are not upgrading their phones for AI features. The figure sits around 10 to 11 percent. People still care more about battery life, cameras, stability, durability, and performance. Chinese manufacturers moved with that insight. Xiaomi, Honor, Vivo, and others doubled down on fundamentals. Large batteries. Silicon carbide charging. Better thermals and even liquid cooling. Efficient open-source models running directly on the device. These improvements are felt every day. So Apple faced a practical decision. Delay another product cycle, or secure a reliable reasoning engine fast. Licensing Gemini became the simplest way to ship a working assistant while maintaining control. This part matters. Gemini does not run on Google servers. It runs inside Apple’s Private Cloud Compute environment, using Apple hardware and Apple isolation. Google never sees the data. Privacy stays in place while capability is delivered. There is also a long-standing financial symmetry here. Google pays Apple close to 20B dollars a year to stay the default search engine. Apple now pays a smaller amount in return for reasoning capability. Both sides get strategic value. The larger question is what this means for the future of AI infrastructure. If Apple can rent the intelligence it needs, then replace it later with its own model, the industry may need to rethink the enormous spending on training and scaling every model internally. Open-source models continue to improve. On-device intelligence is getting stronger. Users reward practical reliability more than parameter counts. This suggests that the real advantage may shift away from owning the biggest model toward owning the execution layer. The layer that turns user intent into real action. Apple is renting today. Once its own model reaches the required level, it can switch away from Gemini without the user noticing. The interface stays familiar. The ecosystem remains Apple’s. The economics shift back to Cupertino. The Apple and Google deal is more than a short-term patch. It is a preview of how intelligence may be delivered going forward: modular, interchangeable, and optimized for integration rather than brute force. The full analysis is in the video above. Would love to hear your perspective on what this means for the future of AI spending and platform strategy. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

5. 

   2 Nov 2025

   The Real Platform Shift: When AI Starts Moving the World

   Humanoid robots are crossing the line between experiment and deployment. 1X’s new NEO may still rely on human tele-operators, but its purpose is clear - to feed the training data that will teach the next generation of autonomous robots how to move, balance, and adapt. Under the hood, this revolution is powered by models like NVIDIA’s GR00T N1 and Tesla’s Optimus - both running vision-based neural networks capable of learning from simulation and real-world video data. This isn’t the “AI boom” we know. It’s the next one. Physical AI - where intelligence leaves the screen and enters the real world. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

6. 

   29 Oct 2025

   The Training Race Is Over. The Power Race Has Begun. Welcome Qualcomm.

   Qualcomm has entered the AI data center race now it unveiled two new rack scale AI inference accelerators the AI 200 & AI250 along with the 200 MW deal announced with the Saudi PIF backed company HUMAIN and like with all announcements AI these days its stock had a pretty good reaction now this isn’t just another GPU launch it’s a sign that the training race is plateauing and the next decade will be won on inference efficiency how many tokens you can serve per MW now for the last five years Nvidia and AMD owned the training era massive HBM stacks terabyte per second links 700 watt GPUs but the bottleneck has shifted training happens once but inference happens millions of times Inference already consumes 70% of all AI compute power and the data center electricity consumption is set to double to about 945 TWh by 2030 so that’s the new economic frontier where cost cooling and power delivery will decide who wins so the AI200 is built precisely for that frontier each accelerator packs 768 GB of LPDDR5X memory inside a 160 kW liquid cooled rack that’s about 4 x the capacity of Nvidia’s H200 so Qualcomm trades bandwidth for capacity and efficiency keeping entire model caches resident on-card instead of streaming data between GPUs so that simple change cuts inference latency by 20 -30% and more importantly the power draw by about 25% so in practice large models like Llama 3 can now run fully resident no sharding no external fetches just lower latency and lower energy it’s the same playbook ARM used against x86 win on performance per watt not brute force hardware alone won’t win Nvidia has CUDA AMD has ROCm so Qualcomm needs a mature AI stack that runs Pytorch ONNX vLLM etc. seamlessly with that Qualcomm could anchor a new class of inference deployments which are smaller localized and energy aware and this is where Qualcomm quietly has an advantage its Arduino acquisition adds billions of edge devices to its ecosystem those boards can run tiny AI models locally then send compressed signals to AI 200 racks for contextual reasoning that’s an edge to core architecture sensors handled detection racks handled reasoning and cloud handles coordination and so if Qualcomm powers lower cost per token cuts data movement and builds the software to match its hardware it could become the efficiency layer of AI quietly powering the shift from cloud scale to energy scale AI! This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

7. 

   26 Oct 2025

   ChatGPT Atlas and the Browser Wars Ahead

   ChatGPT Atlas and the Browser Wars Ahead When OpenAI announced ChatGPT Atlas, it sounded simple at first. Another browser. Another interface. But this launch is much more important than that. It’s OpenAI’s first real attempt to move beyond conversation and into control ,to give ChatGPT a permanent home inside your daily digital life. Why OpenAI needed this For all the hype, ChatGPT’s usage hasn’t evolved much. Most people still use it for the same three things: writing, summarizing, and asking quick questions. OpenAI’s own research shows that around 80 percent of conversations fall into these basic use cases. The app has scale, but not depth. Enterprise adoption is growing, but most companies still treat it as a productivity tool, not infrastructure. That’s why OpenAI is now creating new “surfaces” - places where ChatGPT can live, remember, and act. The ChatGPT App Store was the first step. Atlas is the second. One expands outward through apps, the other moves inward into your browser and workspace. What Atlas can do The Atlas demo showed three big upgrades that quietly redefine how we use the web. Contextual memory It remembers your browsing history, open tabs, and ongoing work. If you’re researching something, it can pick up where you left off. It feels more like a coworker with recall than a search box. On-page assistant You no longer need to copy text into ChatGPT. The assistant lives right on the page. You can highlight text, summarize sections, or ask questions without switching windows. Agentic mode This is the real leap. ChatGPT can now take simple actions - opening tabs, filling forms, comparing data, even building carts, all with your supervision. It’s not just giving answers, it’s performing steps. Comet vs Atlas Perplexity’s Comet browser was the first to experiment with this “agentic” idea. It lets the AI click, scroll, and act inside web pages. It’s fast, flexible, and feels futuristic. Atlas does it differently. It’s slower and more controlled. Every step happens with user permission, memory is optional, and data stays inside a sandbox. The power shift Here’s where it gets interesting. If a browser can read Gmail, Docs, Amazon, or Booking.com directly and make its own decisions, then those sites become utilities. They lose control over discovery. The browser becomes the recommendation layer. That changes the economics completely. Google’s ad business depends on search intent. Amazon’s depends on product discovery. Booking.com’s depends on being the middleman. If the agent is the one choosing what we see, all of that collapses. Search ads, sponsored listings, and paid visibility don’t matter when the AI decides instead of the user. This is the next big shift in power - from platforms to interfaces. And it’s easy to guess how it ends: resistance, lawsuits, and eventually, negotiated partnerships. Every platform that once owned discovery will now have to pay for access to the agent. The new gatekeeper The browser now owns the user. It’s not just a window to the web anymore; it’s the web’s new operating system. If Atlas succeeds, OpenAI won’t just power search - it will mediate how people experience the internet. That’s not a technical change. It’s a redistribution of power. Because whoever owns the agent, owns the audience. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

8. 

   28 Jul 2025

   ChatGPT Agent - What can I actually do with this?

   Recently, I gave ChatGPT's new "Agent" mode a straightforward task: Find a tote bag to gift my wife. Simple enough, right? Here's what actually happened: * It opened up a browser container (good start!). * Attempted to log into Amazon but couldn't get past authentication. * Struggled with dynamically loaded sites - simply couldn't handle modern, JS-heavy pages. * Lost track of the original intent halfway through. * Ended up suggesting obscure luxury handbags priced at around AED 4,000. * After over five minutes of watching it flail, I manually intervened and stopped the process. Just for comparison, I gave GPT-4o the exact same prompt: * Quick web search → clear, relevant results → done in 23 seconds. The contrast was stark. Why is this happening? Despite its impressive-sounding name and sleek UI, ChatGPT Agent tasks still require continuous manual supervision. It often feels like babysitting rather than automating. Instead of true autonomy, I got: * Repetitive loops * Frequent breakdowns on interactive sites * A constant need to step in and redirect or correct the task Here's what I'd ideally want a real "agent" to do: * System-level automations: Ability to run local scripts, manage files, and adjust settings. * Context-aware recommendations: Observing my habits, identifying recurring workflows, and suggesting intelligent automations. * Persistent memory: Remember context and user preferences across multiple tasks and sessions. * Robust error handling: Automatically retry, replan, and recover from failures. * API integration: Reliably switch to APIs when UI-based interactions fail. * Transparency: Clear tracking of state, actions, token usage, and auditability. Conclusion Right now, ChatGPT Agent feels more like a glorified macro in a sandbox not a true autonomous assistant. It's early days, and perhaps expectations should be tempered accordingly. But let's not confuse a slick interface and containerized control with genuine agentic capability. True AI agents must do more than click buttons, they need to think, learn, adapt, and collaborate. We're not there yet. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

9. 

   1 Jun 2024

   🚀 Top 3 AI Takeaways from Google I/O and Microsoft Build 2024! 🚀

   🎥 The key takeaways this time are: AI in Everything 🤖 Agents for Automation🛠️Custom Silicon Wars 🌟 This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit digdeeptech.substack.com (https://digdeeptech.substack.com?utm_medium=podcast&utm_campaign=CTA_1)

Podcast Authority Score: 10 / 100

A composite of feed quality, social presence, YouTube performance and engagement. Read the methodology.

Quality

21

Social presence

0

YouTube

0

Engagement

0

Contact DigDeep Tech Podcast

Guest appearances

Does not typically book guests

Based on episode analysis; this does not confirm that the show is currently accepting guests.