If You Squint, You Can See AGI
The Dawn of "Daily General Intelligence"
On April 16th, 2025, OpenAI unveiled ChatGPT o3, and something clicked for me after using it for the last 3 days. While the tech world debates whether we're approaching "super-intelligence," what's happening is perhaps more immediately transformative: the arrival of what I call daily general intelligence.
OpenAI's benchmarks tell part of the story: o3 demonstrates stronger chain-of-thought reasoning and a context window large enough to process the entirety of The Lord of the Rings. But the real game-changer? The price: approximately $10 per million input tokens and $40 for output—slashing costs dramatically compared to previous offerings.
The competitive landscape makes this evolution even more fascinating. Google's Gemini 2.5 Flash introduced a "hybrid-reasoning" approach with an adjustable "thinking budget" slider. Meanwhile, Anthropic's Claude 3.7 Sonnet offers visible step-by-step "extended thinking," though at premium pricing.
From Research Rabbit Hole to Blogging Bliss
Last week, I needed to upgrade my MacBook Pro M4 setup with two 32" 4K, 120 Hz monitors under $1K each, connected through an OWC 96W Thunderbolt dock. Normally, this would consume an entire evening of research, jumping between spec sheets and Reddit threads.
Instead, I opened ChatGPT o3 and typed possibly the laziest prompt imaginable:
"What monitors should I buy?"
Seven follow-ups later, I was looking at:
A curated short-list (Dell S3225QS, LG 32UQ750) with street prices and warranty notes
A cost/latency comparison of DisplayPort 1.4 versus HDMI 2.1 for 120Hz
An annotated wiring diagram mapping USB-C → DisplayPort 1.4 (dock) and Thunderbolt 4 → DP Alt-Mode (laptop)
Amazon links for a $12 Club3D TB4-to-DP cable and a $9 DP-to-DP run
Total research time? About 15 minutes. The time saved didn't just eliminate a tedious task—it gave me back hours I could spend on what I actually enjoy: writing this blog post.
What Makes This Different
This monitor-shopping experience crystallized what makes o3 feel different from previous AI interactions. While it's not "super-intelligent," it represents a striking step toward what matters in daily life: practical intelligence that saves time on real tasks.
The monitor shopping scenario isn't an isolated example. Last month, I asked o3 to synthesize 20 recent papers on agentic AI frameworks for biopharma workflows. It extracted key concepts, clustered methodologies, highlighted best practices, and generated a presentation-ready radar chart—compressing hours of literature review into a lunch break.
What's happening isn't just incremental improvement. It's the emergence of a tool that consistently collapses hours of cognitive work into minutes, freeing up time for the creative and strategic thinking that humans still do best.
Four Dimensions of Transformation
Looking at today's landscape of flagship AI models, four key dimensions of competition are emerging:
1. Reasoning at the Price of Yesterday's Toy Models
OpenAI's o3 delivers deep reasoning capabilities at approximately $10 per million input tokens and $40 for output. For startups throttling burn rates, that's not just an incremental improvement—it's transformative.
2. "Hybrid" vs. "Always-On" Intelligence
Gemini 2.5 Flash wisely lets teams cap expensive slow-thinking passes with its "thinking budget" slider, but o3's always-on deep reasoning still edges it in factual reliability and creativity—especially when prompts are ambiguous.
3. Claude's Visible Thoughts Are Great—When You Have Time
Claude 3.7's explicit step-by-step traces are pedagogically brilliant, yet the model often runs 2-3× longer and costs ~1.5× more than o3 for the same task. In fast-iterating workflows (coding, customer support), o3's latency wins.
4. 99% Smarter...Still 1% Silly
Despite killer benchmark numbers, o3 occasionally produces paradoxes ("USB-C is faster because it uses fewer electrons") or fabricates nonexistent product SKUs—reminders that the road to true AGI still has potholes.
Why This Matters Now
Three years ago, most roadmaps placed human-level AGI in the early 2030s. Today, o3 can already draft code, design experiments, debug legal clauses, plan vacations, and explain college physics—all for less than the price of a latte per thousand words.
The revelation isn't that we've achieved perfect AGI—we haven't. It's that when a model becomes "good enough" for 99% of day-to-day cognitive tasks, the remaining 1% of weirdness stops being a blocker and becomes merely an amusing footnote.
This shifts the conversation from "when will AGI arrive?" to "how will we redesign our workflows around the intelligence we already have?"
Tomorrow's Playbook, Available Today
For those looking to capitalize on this shift immediately:
Budget for Depth, Not Breadth – Use o3 for tasks requiring deep reasoning; reserve cheaper "flash" modes for high-volume routine operations.
Prompt for Plans, Not Paragraphs – Ask for process-oriented outlines; the model reliably identifies steps you didn't know you were missing.
Instrument Everything – Track latency, cost, and error rates per task type. O3's economics make rigorous A/B testing affordable for everyone.
Keep Humans in the Loop – That 1% silliness will inevitably strike precisely when an unchecked hallucination reaches production.
The Question on the Horizon
If today's "daily general intelligence" is already compressing hours of cognitive work into minutes, how will your organization reimagine workflows when the next iteration eliminates the remaining 1% of errors?
For now, I'm enjoying the extra evening I gained by not researching monitors—and wondering what other tasks I've been doing the hard way all this time.
If you squint just right, you can already see AGI taking shape in our daily lives—not with a dramatic singularity, but through quietly transformative moments like these.