Observations · AI and Business Since 2022
Software development
The most transformed profession of the era. Code completion gave way to agentic tools that plan, write, test, and ship multi-hour units of work. By 2025, AI wrote a large share of new code at major companies, Stack Overflow traffic had collapsed, and entry-level engineering hiring contracted measurably.
What drove it
GitHub Copilot, ChatGPT, Claude (especially Claude Code and the Claude 3.5+ line), Cursor, Devin, OpenAI Codex, and the MCP standard that let agents reach real tools.
The evolution
2022
ChatGPT instantly became a coding companion — within weeks, developers were pasting errors into a chatbot instead of searching Stack Overflow.
- OpenAI launches ChatGPT — the generative AI era begins
It turned years of lab research into a mass-market product overnight and forced every major technology company to respond within weeks. Google declared an internal code red.
2023
Copilot went enterprise, GPT-4 raised the ceiling, the LLaMA leak and Llama 2 opened the model layer — and Stack Overflow's traffic decline made the substitution measurable.
- Microsoft commits ~$10B to OpenAI
It tied the most important AI lab to the second-largest cloud provider and signaled that frontier AI would be financed by Big Tech capital, not standalone startups.
- ChatGPT hits 100M users — fastest-growing consumer app in history
The statistic became the single most-cited proof that generative AI was a mainstream consumer phenomenon, not a tech-insider toy.
- Google rushes out Bard; a demo error erases ~$100B of Alphabet
It crystallized the perception that the AI research leader was reacting defensively, and showed that markets would brutally price AI execution.
- Microsoft ships AI Bing, then puts Copilot inside Office
The first time an incumbent shipped frontier AI directly into products used by billions, and an open assault on Google's core business.
- GitHub Copilot goes enterprise as Stack Overflow traffic slides
Software development became the first profession with measurable, at-scale AI workflow substitution — visible in the decline of the web's canonical programming Q&A site.
- Meta's LLaMA weights leak — open-source AI ignites
It ignited the open-source LLM ecosystem — Alpaca, Vicuna, llama.cpp — and the defining open-versus-closed debate of the era.
- GPT-4: AI passes the bar exam
GPT-4 reset expectations of what AI could do in professional domains — law, medicine, finance, coding — and became the benchmark every rival measured against for over a year.
- Anthropic launches Claude; Claude 2 goes public in July
It established a credible safety-focused rival to OpenAI and made long-context document processing a competitive differentiator.
- China answers: Baidu's Ernie Bot, Alibaba's Tongyi Qianwen
It opened the parallel Chinese LLM race under a distinct regulatory regime, confirming generative AI as a US–China strategic competition.
- NVIDIA joins the trillion-dollar club as the H100 boom begins
It established AI compute as the scarcest strategic resource in technology, and proved the surest profits of the boom were in the hardware layer.
- Llama 2: open weights become a sanctioned Big Tech strategy
Open weights went from leak to strategy, giving every company a credible free alternative to paid frontier APIs.
- Amazon commits up to $4B to Anthropic
It completed the alignment of every frontier lab with a hyperscaler patron — frontier AI economics now required Big Tech capital and compute.
- Mistral 7B: Europe's open-weight challenger arrives
A small European startup shipped competitive models months after founding, proving efficient open weights were a viable strategy against US giants.
- OpenAI DevDay: GPT-4 Turbo, custom GPTs, and the platform play
OpenAI pivoted from model vendor to platform company, aiming to own the application layer the way app stores owned mobile.
- OpenAI's board fires Sam Altman — and reinstates him in five days
The crisis exposed how fragile governance was at the world's most important AI company — and showed that investors, employees, and Microsoft held the real power, not the safety-oriented board.
- Google announces Gemini 1.0, its answer to GPT-4
Google's first credible claimed answer to GPT-4 after a year of being perceived as behind.
2024
The agentic turn: Devin demonstrated autonomous engineering, Claude 3.5 Sonnet became the developer favorite, and open models (Llama 3.1, Qwen 2.5) reached the frontier.
- Gemini 1.5 brings the 1M-token context window
Long context reset expectations for how much information one model could reason over; the image controversy became the year's cautionary tale on guardrail tuning.
- Microsoft partners with Mistral, bringing Mistral Large to Azure
It underscored Microsoft's strategy of backing multiple model providers beyond OpenAI — and drew immediate regulatory scrutiny of Big Tech–AI tie-ups.
- Claude 3: Anthropic reaches the frontier
It made Anthropic a credible frontier competitor and introduced the tiered price-performance lineup that became an industry pattern.
- Devin and the agentic-coding turn
The industry's frame shifted from code completion to autonomous agents that plan and execute whole tasks.
- NVIDIA announces Blackwell at GTC
It set the hardware roadmap for trillion-parameter models and reinforced NVIDIA's grip on AI infrastructure.
- Meta releases Llama 3
It cemented Meta's leadership of the open-weight ecosystem and kept a high-quality free alternative pacing the closed labs.
- GPT-4o: real-time multimodal AI goes free
Real-time, natural multimodal interaction reached a mass-market free product, resetting the price-performance bar.
- AI goes on-device: Apple Intelligence and Copilot+ PCs
Generative AI became a default operating-system feature on billions of devices, with privacy-preserving local inference as the selling point.
- Claude 3.5 Sonnet and Artifacts: better, faster, cheaper — with a workspace
Proof of how fast price-performance was improving, and a new interaction model built around producing real work rather than chat.
- Llama 3.1 405B: the first open-weight frontier model
It collapsed the assumed gap between open and closed AI, changing the build-versus-buy calculus for every company.
- xAI builds Colossus — 100,000 GPUs in 122 days
It demonstrated how fast a new entrant could stand up frontier-scale compute and credibly challenge incumbents.
- The reasoning era: OpenAI ships o1, then announces o3
Inference-time reasoning emerged as a new scaling axis distinct from model size — capability could now be bought with thinking time.
- Alibaba's Qwen 2.5: China takes open source seriously
A turning point where Chinese open models began rivaling — and drawing developers away from — Western options like Llama.
- OpenAI raises $6.6B at a $157B valuation
It nearly doubled OpenAI's valuation in months and tied its corporate structure to its capital needs — a thread that defined 2025.
- Amazon doubles down: $8B total into Anthropic
Deepened the hyperscaler–lab alliance pattern and advanced Amazon's bid to challenge NVIDIA with custom AI chips.
- DeepSeek-V3: frontier-class performance at a fraction of the cost
It challenged the assumption that frontier models required hundreds of millions in training spend, setting up January's R1 shock.
2025
Claude Code, MCP, and vibe coding made agents the default workflow; the Windsurf bidding war priced the category; AI began writing a large share of new production code.
- Stargate: a $500B bet on American AI infrastructure
The largest infrastructure commitment in AI history. Frontier AI officially became national industrial policy.
- Operator: the year of agents begins
The industry narrative shifted from chatbots that answer to agents that act.
- The DeepSeek shock: R1 wipes ~$600B off NVIDIA in a day
Frontier-level reasoning, cheap and open, directly challenged the assumptions underneath hundreds of billions of dollars in US AI capex.
- Grok 3: brute-force compute reaches the frontier
A two-year-old startup reached the frontier through sheer compute scaling, validating the bigger-clusters-win thesis.
- Claude 3.7 Sonnet and Claude Code: hybrid reasoning meets the terminal
Hybrid reasoning became the industry pattern, and a frontier lab put its own agentic coding product directly into developers' hands.
- Gemini 2.5 and Veo 3: Google returns to the front
After two years trailing OpenAI, Google led on both reasoning quality and generative media in the same season.
- MCP becomes the industry standard for connecting AI to tools
For the first time, all major rival labs aligned on one open standard — the plumbing the agent era required.
- Claude 4: long-horizon autonomy becomes the competitive axis
Long-horizon autonomy — not chat quality — became the new axis of frontier competition.
- Meta pays $14.3B for Scale AI and starts the talent war
The most dramatic talent war in tech history; Meta chose to buy its way back to the frontier.
- Vibe coding goes mainstream; the Windsurf saga rewrites AI M&A
AI coding tools proved to be generative AI's first killer app — valuable enough to trigger billion-dollar bidding wars.
- China's open-weight wave: Qwen 3, Kimi K2, GLM-4.5, DeepSeek V3.1
China became the center of gravity for open-source AI — frontier-adjacent capability at a fraction of US API prices.
- GPT-5 becomes the default for ChatGPT's hundreds of millions
Hundreds of millions of free users got their first reasoning model; the launch's incremental feel also fed the year's bubble debate.
- Anthropic triples its valuation to $183B in six months
It validated the enterprise-first, coding-led strategy as a counterweight to OpenAI's consumer scale.
- The AI browser war: ChatGPT Atlas and Perplexity Comet
The browser — the gateway to the web's ad economy — became the new battleground for AI distribution.
- AI reaches the workforce: Amazon cuts 14,000 as the evidence mounts
AI-driven workforce reduction moved from prediction to documented practice, with major employers explicitly linking headcount to AI.
2026
Frontier flagships (GPT-5.5, Gemini 3.5, Claude Fable 5) compete on operating software end-to-end; Apple made the model a swappable OS primitive — Claude, Gemini, and OpenAI behind one Swift interface, agents inside Xcode; China's open stack (Qwen, DeepSeek V4 on Huawei silicon) sets the price floor; and DiffusionGemma put a non-autoregressive generation path into open, deployable form.
- OpenAI signs a $10B inference deal with Cerebras
A deliberate diversification away from NVIDIA-only infrastructure, validating specialized inference silicon as a category.
- The AI layoff wave: payroll becomes capex
The first sustained period in which profitable companies explicitly tied workforce reductions to AI adoption and capex funding.
- Qwen nears a billion downloads — and closes its flagship weights
Chinese open models became the global default — while even open-source champions began holding back frontier weights for commercial advantage.
- DeepSeek V4 runs on Huawei silicon: the decoupling milestone
China's frontier AI development visibly decoupled from US hardware: a top open model paired with a domestic compute stack for the first time.
- GPT-5.5: OpenAI's flagship becomes an agent
OpenAI's flagship line moved decisively from chat toward computer-operating agents — with reliability, not capability, as the headline metric.
- Cerebras IPO: biggest US tech listing since Uber, up 68% on debut
The first blockbuster pure-play AI chip IPO of the cycle, testing public appetite for NVIDIA challengers.
- Google I/O 2026: Gemini 3.5 and Gemini Omni
Google's frontier stack — intelligence with action — pushed across consumer search, developer tools, and enterprise agents in one coordinated move.
- MCP drops sessions — the agent protocol becomes ordinary plumbing
Every one of these changes is boring: load balancing, standard auth, trace propagation, a published deprecation window. That is exactly the point. Protocols stop being demos when they acquire the unglamorous properties that let ordinary operations teams run them — and the value migrates off the model and onto the interfaces and workflows built around it.
- Anthropic raises $65B at $965B, overtakes OpenAI, files for IPO
The world's most valuable AI startup, and the first frontier-lab IPO filing — a watershed for public-market AI exposure.
- Apple makes the model layer a swappable OS primitive
It commoditizes the model layer at the level of the operating system. A provider-neutral interface baked into the platform SDK turns the model into an interchangeable component and makes the surrounding workflow — distribution, integration, trust — the durable asset.
- OpenAI confidentially files a draft S-1
Both leading frontier labs entered the public-markets pipeline within two weeks. That turns 'private AI valuations' into an imminent question of public disclosure: audited financials, named risk factors, and the first comparable look at frontier-lab economics.
- Claude Fable 5: the first broadly available Mythos-class model
A new top capability tier reached the open market days after Anthropic's record raise, with safety gating presented as the unlock.
- Google releases DiffusionGemma, an open text-diffusion LLM
It makes a non-autoregressive generation path real, open, and deployable on mainstream serving infrastructure for the first time — a different lever on inference cost and speed than building a bigger model. Whether the paradigm holds is unproven: quality lags and adoption is unknown.
- Mastercard launches Agent Pay for Machines
The hard problem in agentic commerce is not moving money — it is proving an agent's identity and constraining its authority. Putting that identity, consent, and permission layer on shared infrastructure is the trust standard the agent economy was missing.
- Salesforce buys Fin for $3.6B on a 76% resolution claim
The largest consolidation yet in AI customer support, and the number at the centre of it deserves reading carefully. A resolution rate measures volume that never reached a human. It does not measure whether the customer's problem was solved, whether they came back three days later, or whether they simply gave up. Those are different questions, and only the first one is easy to instrument — which is precisely why it is the one that gets quoted.
Key sources
The papers, announcements, and reports that document this industry's shift. Each also lives in the Library.
What businesses do differently
Teams budget for AI tooling per engineer the way they once budgeted for cloud. Senior engineers spend more time reviewing and specifying, less time typing. 'Vibe coding' made prototyping nearly free, so the bottleneck moved to judgment: what to build and what to trust.
What this means for founders and leaders
Software is cheaper to produce and easier to copy than ever. Velocity is table stakes; durable advantage comes from distribution, data, and taste. If your engineers aren't using agentic tools daily, you are paying a competitor's payroll.