Uploaded January 2026 | Updated September 2026, 2 weeks ago
Most people assume AI “remembers everything” — every chat, every command, every conversation. But that’s not how today’s systems actually work. On this episode of Today in Tech, Keith Shaw talks with Manifest AI CEO Jacob Buckman about how AI memory really works under the hood, why chatbots feel so different from humans, and what has to change for true long-running digital agents to become reality.
Jacob explains concepts like short-term vs. long-term AI memory, context windows, KV caches, and “scratchpad” summaries in plain language. He uses analogies from medicine and the movie Memento to show why current AI tools can ace a single conversation but struggle to stay on task over hours, days, or projects. They also dig into hallucinations, why simply “making models bigger” isn’t enough, and how new architectures like power retention aim to give AI a more human-like ability to remember what actually matters over time.
You’ll learn:
* Why AI remembers everything inside a chat window but almost nothing between sessions
* How today’s memory tricks (summaries, scratchpads, huge context windows) still fall short
* How memory limits hold back reliable AI agents for coding, research, and creative work
* Why better long-term memory could cut hallucinations and boost trust in business use cases
* What “power retention” is — and how it could reshape the next generation of AI systems
Chapters:
00:00 – Do AIs really remember everything? (cold open)
00:24 – Welcome to Today in Tech + guest intro
00:46 – What people get wrong about AI memory
01:39 – How chatbots “remember” inside a single conversation
02:19 – Human vs. AI memory: what carries over between chats
03:18 – KV cache, context, and how AI “photographic memory” works
04:11 – Why AIs only keep short natural-language notes between sessions
05:18 – The doctor analogy: AI as a professional reading their notes
06:15 – Human short-term vs. long-term memory vs. AI’s two-track system
07:47 – Why current AIs struggle with long-running conversations and tasks
09:18 – The Memento analogy: AI as a note-writing protagonist
10:11 – Why people treat AI like search—and where that breaks down
11:19 – Limits of agents: staying on task over complex, multi-step work
11:59 – Coding agents, debugging, and why they still fall short
13:21 – Long tasks, losing focus, and where memory becomes the bottleneck
13:44 – Are hallucinations a separate problem—or just another kind of error?
15:26 – Scaling, research, and how to reduce AI mistakes
16:00 – Simple lookups vs. deep research tasks
16:39 – Why long, multi-step research still breaks today’s agents
17:39 – Is long-term memory the biggest bottleneck now?
18:44 – Building trust with AI like you would with a new employee
20:09 – Why people over-trust AI as a neutral “superjudge”
21:27 – How businesses should roll out AI safely and gradually
22:06 – What big labs are doing about AI memory today
22:57 – Context windows and the push for massive short-term memory
24:25 – Scratchpads, summaries, and today’s long-term memory hacks
25:12 – Why Jacob thinks we need new architectures
25:38 – Introducing “power retention” vs. traditional attention
26:33 – Why perfect photographic memory is not actually ideal
27:22 – Human forgetting as compression — and what AI can learn from it
27:50 – How power retention remembers only what matters
28:38 – Scaling powerful memory across an AI’s entire experience
29:27 – What better memory unlocks for agents and complex tasks
30:16 – Intelligence per dollar and why cost matters for memory
31:32 – Accuracy vs. creativity: does creative AI need long-term memory too?
32:23 – Creative workflows, iteration, and why AI keeps “re-breaking” designs
34:01 – From helper to owner of the long-term vision in creative projects
34:57 – Could AI fully drive big creative or product journeys?
36:08 – Will there be an “AI format war” over architectures?
37:28 – Infrastructure, inference platforms, and adoption friction
38:20 – Attention vs. retention: who becomes the backbone of future models?
39:01 – Rewriting Memento if the protagonist had future AI
40:30 – Flipping the roles: AI as long-term planner, human as short-term executor
41:25 – Coming full circle: remembering Memento and AI memory
41:41 – Closing thoughts + what’s next for AI memory research
#todayintech #keithshaw
Follow TECH(talk) for the latest tech news and discussion!
----------------------------------
Keith Shaw
linkedin.com/in/shawkeith
SUBSCRIBE: youtube.com/@CWTECHtalk
TIKTOK: tiktok.com/@todayintechpodcast
Most people assume AI “remembers everything” — every chat, every command, every conversation. But that’s not how today’s systems actually work. On this episode of Today in Tech, Keith Shaw talks with Manifest AI CEO Jacob Buckman about how AI memory really works under the hood, why chatbots feel so different from humans, and what has to change for true long-running digital agents to become reality.
Jacob explains concepts like short-term vs. long-term AI memory, context windows, KV caches, and “scratchpad” summaries in plain language. He uses analogies from medicine and the movie Memento to show why current AI tools can ace a single conversation but struggle to stay on task over hours, days, or projects. They also dig into hallucinations, why simply “making models bigger” isn’t enough, and how new architectures like power retention aim to give AI a more human-like ability to remember what actually matters over time.
You’ll learn:
* Why AI remembers everything inside a chat window but almost nothing between sessions
* How today’s memory tricks (summaries, scratchpads, huge context windows) still fall short
* How memory limits hold back reliable AI agents for coding, research, and creative work
* Why better long-term memory could cut hallucinations and boost trust in business use cases
* What “power retention” is — and how it could reshape the next generation of AI systems
Chapters:
00:00 – Do AIs really remember everything? (cold open)
00:24 – Welcome to Today in Tech + guest intro
00:46 – What people get wrong about AI memory
01:39 – How chatbots “remember” inside a single conversation
02:19 – Human vs. AI memory: what carries over between chats
03:18 – KV cache, context, and how AI “photographic memory” works
04:11 – Why AIs only keep short natural-language notes between sessions
05:18 – The doctor analogy: AI as a professional reading their notes
06:15 – Human short-term vs. long-term memory vs. AI’s two-track system
07:47 – Why current AIs struggle with long-running conversations and tasks
09:18 – The Memento analogy: AI as a note-writing protagonist
10:11 – Why people treat AI like search—and where that breaks down
11:19 – Limits of agents: staying on task over complex, multi-step work
11:59 – Coding agents, debugging, and why they still fall short
13:21 – Long tasks, losing focus, and where memory becomes the bottleneck
13:44 – Are hallucinations a separate problem—or just another kind of error?
15:26 – Scaling, research, and how to reduce AI mistakes
16:00 – Simple lookups vs. deep research tasks
16:39 – Why long, multi-step research still breaks today’s agents
17:39 – Is long-term memory the biggest bottleneck now?
18:44 – Building trust with AI like you would with a new employee
20:09 – Why people over-trust AI as a neutral “superjudge”
21:27 – How businesses should roll out AI safely and gradually
22:06 – What big labs are doing about AI memory today
22:57 – Context windows and the push for massive short-term memory
24:25 – Scratchpads, summaries, and today’s long-term memory hacks
25:12 – Why Jacob thinks we need new architectures
25:38 – Introducing “power retention” vs. traditional attention
26:33 – Why perfect photographic memory is not actually ideal
27:22 – Human forgetting as compression — and what AI can learn from it
27:50 – How power retention remembers only what matters
28:38 – Scaling powerful memory across an AI’s entire experience
29:27 – What better memory unlocks for agents and complex tasks
30:16 – Intelligence per dollar and why cost matters for memory
31:32 – Accuracy vs. creativity: does creative AI need long-term memory too?
32:23 – Creative workflows, iteration, and why AI keeps “re-breaking” designs
34:01 – From helper to owner of the long-term vision in creative projects
34:57 – Could AI fully drive big creative or product journeys?
36:08 – Will there be an “AI format war” over architectures?
37:28 – Infrastructure, inference platforms, and adoption friction
38:20 – Attention vs. retention: who becomes the backbone of future models?
39:01 – Rewriting Memento if the protagonist had future AI
40:30 – Flipping the roles: AI as long-term planner, human as short-term executor
41:25 – Coming full circle: remembering Memento and AI memory
41:41 – Closing thoughts + what’s next for AI memory research
#todayintech #keithshaw
Follow TECH(talk) for the latest tech news and discussion!
----------------------------------
Keith Shaw
linkedin.com/in/shawkeith
SUBSCRIBE: youtube.com/@CWTECHtalk
TIKTOK: tiktok.com/@todayintechpodcast
