What My Algorithm Read This Week ( Feb 22, 2026)
I built a system that curates my own newsfeed — not optimized for engagement or ad revenue, just for what’s actually useful to my work. Every week, I’m sharing what it surfaces. 35 feeds scanned, 100+ items analyzed, 4 themes identified.
The Junior Developer Paradox
Multiple sources converging on a counterintuitive finding about AI and entry-level talent.
The prevailing narrative says AI will eliminate junior developer jobs first. This week, the data says otherwise.
IBM is tripling entry-level hiring after “finding the limits of AI adoption.” Not cutting juniors — adding them.
Thoughtworks puts it bluntly: “Juniors are more profitable than they have ever been. AI tools get them past the awkward initial net-negative phase.”
And Boris Cherny: “Someone has to prompt the Claudes… Engineering is changing and great engineers are more important than ever.”
The pattern: AI doesn’t replace juniors — it accelerates their path to productivity. Companies that assumed AI would reduce headcount are discovering they still need humans to direct, validate, and coordinate.
Worth reading alongside Martin Fowler on task switching in supervisory programming — even with AI doing the execution, humans managing AI agents face real cognitive overhead.
The Epistemic Crisis
We’re building systems we don’t understand — and several pieces this week grapple with what that means.
What Is Claude? Anthropic Doesn’t Know, Either — The New Yorker’s long-form piece on the fundamental uncertainty at the heart of AI development. If you read one thing this week, this is it.
Cognitive Debt — Margaret-Anne Storey introduces a concept that deserves wide adoption: AI shifts concern from technical debt (code we don’t understand) to cognitive debt (systems whose behavior we can’t predict).
Nick Bostrom on optimal timing for superintelligence — an academic framing of the when-to-deploy question under uncertainty.
OpenAI ships Lockdown Mode — enterprise security features that quietly acknowledge prompt injection is a real, unsolved problem.
The pattern: The gap between capability and comprehension is widening. OpenAI ships physics discoveries while Anthropic admits they don’t fully understand Claude. Security features acknowledge attack surfaces that fundamentally can’t be eliminated.
The tension: the same feeds also report GPT-5.2 deriving new physics results. Capability is advancing even as understanding lags.
The Open Source Velocity Story
Three Months of OpenClaw: 10,000 commits from 600 contributors, 196,000 GitHub stars — in under 90 days. From first commit to category-defining project in a single quarter.
MiniMax M2.5 matches Opus — a Chinese open model claims 80.2% on SWE-Bench Verified. The gap between closed and open models continues to narrow.
Data Engineering Book — open source, community-driven technical guides emerging faster than publishers can commission them.
The pattern: Open source AI development is moving at unprecedented velocity. The question isn’t whether open models will catch up — it’s how fast.
The Canadian angle: Cohere hits $240M ARR, showing enterprise-focused, Canadian-headquartered AI companies finding their niche between open source and the frontier labs.
Platform Power Shifts
Where AI disrupts existing power structures — and where it doesn’t.
Why OpenAI Should Build Slack — the argument that AI labs should own the collaboration layer, not just the model.
Aggregators and AI — Ben Thompson on whether AI breaks aggregation theory.
AI is Going to Kill App Subscriptions — consumer software business models under pressure.
OpenAI’s mission statement evolution — tracked through IRS filings. The word “safely” was deleted. Organizational mission drift, documented in real time.
The pattern: AI is reshaping platform economics. Aggregators that controlled distribution now face models that can generate supply. But the frontier labs themselves are also drifting — and if AI labs become the new aggregators, do the same criticisms apply?
Threads Worth Following
News Publishers vs Internet Archive — AI training debates claiming collateral damage
EU Bans Infinite Scrolling — regulatory action on attention-capture patterns
Smart Sleep Mask Security Disaster — IoT device broadcasting brainwaves to an open MQTT broker
Breaking the Spell of Vibe Coding — fast.ai on the dangers of undirected AI coding
Alberta Angle
Platform Calgary Community Connect — Feb 19 networking event
BBVA Innovation Expo in Edmonton — Dragons Den-style pitch competition
Photonic Quantum Teleportation — BC company + Telus claiming a breakthrough
This digest is generated by my own curation algorithm — 35 feeds, thematic clustering, no engagement optimization. The system is part of an experiment in prompt learning I’m building in the open.