Is computer programming still the main area where AI has made an impact?

The speaker argues that programming is the most pronounced application yet and deserves its own category separate from other non-LM AI advances, though it may not be the sole driver of change in the near term.

What does the video say about AlphaFold and similar biotech advances?

AlphaFold is acknowledged as a landmark achievement but is noted as predating the recent large language model investments; it is described as a tool with its own trajectory, not a direct result of the current boom.

How does AI affect customer support according to the talk?

It’s described as capable of handling basic, scripted conversations and first-line support, reducing the burden, though full replacement and pace of adoption remain uncertain.

Are the AI breakthroughs from the recent investments as game-changing as hype suggests?

The speaker argues that while there are impressive demos and some capabilities, the concrete changes are still narrow and not the massive, broad revolution some booster claims imply; progress has been slower and more incremental.

What is the overall takeaway about the impact of the AI investments?

Overall, the talk concludes that there have been useful, narrow applications but not a handful of 'home runs' yet; the effect is significant but still slower and more limited than some narratives suggest.

Dangerous Question: Has AI Been A Disappointment So Far? | Cal Newport

Deep Questions with Cal Newport

People & Blogs4 min read14 min video

Jan 29, 2026|21,270 views|650|220

Cal Newport Deep Work Deep Life Deep Questions TimblockPlanner Deep Questions Podcast cal newport interview cal newport podcast social media detox productivity tips cal newport productivity cal newport motivation

Save to Pod

Key Moments

TL;DR

AI bets bring modest, narrow gains; programming remains the clearest win, not a revolution.

Key Insights

Most tangible effects from recent AI investments are incremental and outside core programming tasks.

Programming remains the clearest, most impactful application area tied to large language models.

Many claimed breakthroughs are not unique to the latest AI boom; progress is a slow, continuous stream.

Everyday productivity tasks (summarization, report generation, slide creation) show usefulness but limited disruption.

Non-LLM AI developments (e.g., AlphaFold) predate the current wave and shouldn’t be conflated with the latest investments.

There are real risks (misinformation, deepfakes) that accompany the hype, requiring critical scrutiny of claims.

INTRODUCTION: SETTING THE QUESTION AND METHOD

The video opens with a provocative question: given the enormous money pouring into generative AI, what has actually changed in the real world? Cal Newport proposes a disciplined method: examine a broad Reddit discussion from a few months prior, filter out obvious non-LLM results, and separate pure programming advances from other AI progress. He emphasizes two ground rules: treat only effects arising from the latest large language model–driven investments, and isolate programming as its own category because it benefits most directly from the current wave. The exercise aims to separate hype from tangible, first-order changes, acknowledging that some gains—like pattern detection, information summarization, and rapid content production—are real but not earth-shattering. This framing sets the tone for a measured evaluation rather than a panic-driven revolution narrative.

WHAT COUNTS AS A BREAKTHROUGH? A GROUND-LEVEL TEST

A central tension in the discussion is what constitutes a true breakthrough versus routine progress. Some participants imagine advanced reasoning tasks—like analyzing complex laws for constitutional risk or forecasting governance outcomes—as plausible new capabilities from today’s AI. Others push back, arguing that current language models lack genuine imagination or robust scenario planning. The speaker notes that several claimed breakthroughs actually rely on technologies or data pipelines that predate the latest boom, or on other AI modalities, not the core LLM investments. This debate matters because it reframes expectations: if breakthroughs are not uniquely tied to recent billions, the narrative of a sudden, sweeping revolution weakens, and a slower, more patchwork pace emerges.

PATTERN FINDING AND MASS-PIPELINE OUTPUT: NON-LANGUAGE GAINS

A recurring theme is that LLMs excel at pattern finding within large, messy data and can generate outputs at scale, which translates into practical but not revolutionary benefits. Examples include rapid summarization of documents, distilling long texts into bullet points, and producing bite-sized content from vast notes. The thread also highlights slide production as a concrete time-saver: converting a multi-page document into a presentation in minutes. These capabilities are valuable for productivity, but they tend to be framed as improvements in efficiency rather than evidence of a radical, world-changing leap. This section grounds the discussion in concrete, replicable tasks rather than abstract potential.

ALPHAFOLD AND NON-LANGUAGE MODEL ADVANCES: WHAT ABOUT PREEXISTING MILESTONES?

The conversation underscores that some of the most celebrated breakthroughs—like Alphafold’s protein structures—are not products of large language model investments and, in some cases, predate the current AI boom. Alphafold’s success is a landmark in its own right, but it arises from a different lineage of AI research, with distinct data, architectures, and breakthroughs. This distinction matters because it challenges the narrative that every significant AI milestone today is a direct result of the latest LLM-driven funding spree. Recognizing these cross-cutting advances helps prevent conflating diverse AI trajectories into a single monolithic “AI revolution.”

BUSINESS-FOCUSED USES: FROM SUPPORT CHAT TO BORING TEXT AT SCALE

The Reddit thread surfaces several practical, everyday uses that businesses already exploit or could readily deploy: first-level customer support agents handling scripted interactions, summarizing long documents for executives, drafting routine self-reviews, and translating or simplifying dense materials. These tasks, while invaluable for efficiency, are not dramatic breakthroughs but steady improvements in automation and information handling. The discussion notes these capabilities are enabled by current models, yet they rarely rewrite business models or reshape industries on their own. They do, however, reduce time spent on repetitive work and improve throughput for knowledge workers.

HYPE, RISK, AND THE LIMITS OF CURRENT DISRUPTION

A skeptical strand in the conversation warns against overhyping the immediate impact: there are many things AI can do well, but few ‘home runs’ that redefine sectors overnight. The potential for deepfakes, misinformation, and manipulation is acknowledged as a real danger that accompanies the hype. The speaker also dismantles the myth of thousands of secret, breakthrough projects behind corporate walls, arguing that most truly transformative work is incremental, transparent, or already in development long before this latest wave. The overall message is cautious: expect measured gains and realistic timelines rather than a sudden, universal disruption.

LOOKING AHEAD: WHAT MIGHT REALLY CHANGE AND WHEN

Toward the end, the speaker invites a sober forecast: if the current batch of improvements feels incremental, what could trigger a true pivot in the near future? The consensus in the discussion is that the most reliable progress will appear as improvements to narrow, well-defined tasks—documentation, analysis, content creation, and data-driven insights—rather than a wholesale rewrite of human capability. The next wave may hinge on better integration with human workflows, more reliable evaluation, and safeguards against misuse. In short, the trajectory is forward but not instantly transformative; the pace and shape of disruption remain uncertain.

Mentioned in This Episode

●Tools & Products

●People Referenced

AI Content Practical Cheatsheet

Practical takeaways from this episode

Do This

Use AI to summarize long documents quickly

Leverage AI to create slides from text or notes

Automate routine note-taking and documentation

Employ first-line AI chat agents with human oversight for support

Avoid This

Don’t assume AI provides deep, novel insights without human validation

Don’t rely on AI for critical decision-making or legal/ethical judgments

Common Questions

The video catalogs several concrete uses: pattern finding in large data sets, rapid slide production, text summarization, first-line customer support, automatic boring-text writing, and easier distillation of complex text. Timestamped references illustrate how these functions are being realized.