Research rigor, delivery muscle, boardroom judgment, frontier fluency: woven into one outcome-driven strategy for each specific business problem. Grounded in computing, business, and applied finance; sharpened in the field.
Expertise 01
Research craft
Signals
Weak signals today, headlines tomorrow: essays, storyboards, and market calls in every medium. All views personal.
Market Calls
AG Radar: My Market Calls, Scored in the Open
Eight falsifiable market calls with deadlines, kill conditions, and quarterly verdicts. A living board, updated every sweep. The misses stay up.
September 2026 · 6 min read
GenAI x Research
I'm Not a Developer. I'm Building a Macro-Intelligence Dashboard Anyway.
A storyboard from the pilot: turning free government and macro data into one dashboard by directing AI coding agents like an analyst team.
Anyone can claim foresight after the outcome. This board removes that option.
Eight falsifiable market calls, made in July 2026: each one a single sentence with a subject, a measurable threshold, and a deadline, frozen the day it was published. Every call states in advance what would prove it wrong.
02 · The board today
8 calls on the boardAll WatchingFirst quarterly sweep October 2026Board updated September 2026
The board is live: the sweep rotates, and each verdict ring redraws with every pass. Core colour is my conviction going in: purple for high conviction, orange for 60/40, pink for contrarian, white for the deliberate stretch. The ring is the verdict, sky blue while watching, and it will turn green, yellow or red as calls resolve. Hover any blip for its status. Every quarterly sweep updates this post.
03 · The calls
Agentic Big Four60 / 40Watching
Resolves by December 2028 · 6.3% of runway elapsed
By December 2028, at least one Big Four network will publicly market a named subscription or platform offer for recurring tax, risk or finance work in which AI agents are described as the standard delivery workflow and human experts are positioned primarily as reviewers, exception handlers or escalation points. My employer is excluded from scoring this call.
Revenue Up, Headcount FlatHigh convictionWatching
Resolves by FY2028 reporting · 6.3% of runway elapsed
By fiscal-year results covering periods ending no later than December 2028, at least one global consulting or professional-services company above US$10 billion in revenue will report two consecutive years of revenue growth with flat or lower headcount, with management attributing part of the leverage to AI-enabled delivery. My employer is excluded from the qualifying set.
AI-Native ResearchHigh convictionWatching
Resolves by December 2027 · 10.6% of runway elapsed
By December 2027, at least one of Gartner, Forrester, IDC, Omdia or Everest Group will position a conversational, AI-native interface to its licensed intellectual property as the lead paid-access product in its enterprise commercial packaging.
Research IP to Agents60 / 40Watching
Resolves by December 2027 · 10.6% of runway elapsed
By December 2027, at least one of Gartner, Forrester, IDC, Omdia or Everest Group will commercially launch a secure MCP server, enterprise connector or equivalent governed API that lets customer-controlled copilots retrieve its licensed research inside enterprise workflows.
Outcome-Priced EngineeringStretchWatching
Resolves by December 2028 · 6.3% of runway elapsed
By December 2028, at least one engineering and R&D services provider above US$500 million in revenue will publicly make outcome-linked commercial terms a standard contracting option within a named AI-led engineering offer, with fees tied to measurable engineering KPIs.
Hyperscaler Power CapacityHigh convictionWatching
Resolves by September 2027 · 12.8% of runway elapsed
By September 2027, at least one hyperscaler will publicly fund, co-invest in or reserve for multiple years a dedicated production line or defined tranche of manufacturing capacity directly with a transformer or medium- or high-voltage switchgear manufacturer.
AI Load in Grid PolicyContrarianWatching
Resolves by June 2028 · 7.9% of runway elapsed
By June 2028, at least three G20 governments other than the United States and United Kingdom will adopt a statutory, regulatory or cabinet-approved grid-connection or strategic power-equipment manufacturing measure that explicitly cites AI or datacentre load growth.
EREVs Beyond China60 / 40Watching
Resolves by September 2027 · 12.8% of runway elapsed
By September 2027, at least two additional mass-market OEM groups headquartered outside China, excluding Hyundai Motor Group and Stellantis, will have launched, homologated or formally announced an extended-range electric vehicle for a named major market outside China.
04 · The rules
Signal sentences are never edited after publication. Verdicts change only at quarterly sweeps: Playing out, Early, or Wrong, and the wrong calls stay on the board permanently, because a scoreboard with no losses is a marketing page. Calls draw only on public data, exclude my employer from scoring and evidence, and constitute no investment advice. All views are personal.
GenAI x Research
GenAI x Research
I'm Not a Developer. I'm Building a Macro-Intelligence Dashboard Anyway.
Albin George · July 2026 · ~4 min read
The blueprint: the dashboard view this pilot is building toward.
01 · What
A macro-intelligence dashboard in the making, built on data that is free to everyone.
The project: collect the government and macroeconomic sources every strategist already uses, and turn them into one view that is easy to read and easy to extract from. The twist: I am not a developer. I am building it by directing two AI coding agents, Claude Code and Codex, the way I would direct an analyst team.
A few of the sources I have leveraged for the initial pilot:
MoSPIRBI DBIEdata.gov.inWorld BankIMFFRED
02 · Why
Official data is high-value but scattered and revision-prone. The real cost of tracking it manually is not effort. It is inconsistency: a base-year revision missed here, a provisional figure treated as final there.
The old workflow
Bookmarked portals. Quarterly downloads. An Excel file rebuilt by hand, with every revision cycle risking silent errors.
What I wanted
One pipeline that collects, cleans, validates, and refreshes itself on each source's release calendar.
03 · How · Step 1: Encode the SOP
Both agents read standing instructions before doing any work: CLAUDE.md for Claude Code, AGENTS.md for Codex. Most people store coding conventions there. Mine holds a research SOP, the same document I have written for analyst teams for ten years.
CLAUDE.md, research SOP (excerpt)
# Data conventions the agent must never forget
- MoSPI series: FISCAL year (Apr to Mar). World Bank: CALENDAR year
- Never mix ₹ crore with USD million. Convert and label units
- Figures marked "P" are provisional. Re-pull when revised estimates land
- Respect base-year revisions. Never splice old and new series silently
- After any data write: run scripts/validate.py
Written once. Applied on every run. Never reminded.
03 · How · Step 2: Delegate the mess
New sources arrive messy: inconsistent headers, odd formats, patchy documentation. Read all of that in the main session and the agent's planning visibly degrades, the way a steering committee degrades when handed raw field notes. Subagents solve it.
Main agentholds the schema & plan, stays clean
→
Subagentisolated context reads the mess, inventories fields & quirks
→
Returnsonly a clean summary
03 · How · Step 3: Gate the quality
Researchers are professionally paranoid about silent data corruption. Claude Code's hooks fire automatically after the agent acts, so my validation script runs after every data write, and a failure blocks the step.
Stepagent writes data
→
Validation hookrow counts · date continuity · unit sanity
→
Gatepass → proceed · fail → blocked, bounced back
03 · How · Step 4: Put it on the calendar
Government data follows a release calendar, and the pilot pipeline is being wired to follow it. Both agents run non-interactively: Claude Code headless, Codex through its exec command. A scheduled job handles the rest.
Triggereach source's release calendar
→
Scheduled runheadless Claude Code · codex exec
→
Resultre-pull, validate, flag anomalies for review
03 · How · Step 5: Pair the agents
The underrated pattern: use them together. One builds, one reviews, and approval modes start at "propose everything" and earn autonomy, exactly how you would onboard a junior analyst.
Claude Code · the builder
CLAUDE.md context · subagents for isolation · hooks for QC · headless runs.
Codex · the reviewer
A separate review agent inspects changes before commit · graduated approval modes.
Both speak MCP (Model Context Protocol): data connectors built once, shared between them.
04 · Where the pilot stands
A pipeline taking shape
Early runs refresh on release calendars and flag anomalies for review, replacing my quarterly ritual step by step.
Trust, engineered in
Validation gates are proving they can block bad writes before they spread. Not hoping the agent remembered: knowing it cannot proceed otherwise.
An SOP that executes itself
Ten years of research discipline, encoded once in a context file, applied on every pilot run.
A transferable skillset
Brief clearly, isolate deep dives, review summaries. Managing agents turned out to be a skill research leaders already have.
The judgment layer is still the analyst's job. Now it is directly executable.
Next in the pilot: the visualization layer, doubling as my Power BI and Tableau practice.
