Mid-2026 Track Record: 156 Predictions on the Record, 81% Verified

The Zanii Research program is now at 156 verifiable calls. Through end of Q1 2026 the standing record is 127 verified, 18 partial, 11 wrong. An 81% strict verified rate. 93% directional. The distribution is closer to our target band than the 2025 record was, which we read as confirmation that the deployment-versus-capability distinction we introduced in 2025-W49 is calibrating us correctly.

Below: the H1 2026 grading, the two new wrong calls, and the read on the rest of the year.

The 2026 Q1 grade

Of the 24 calls published 2026-W01 through 2026-W24, six are graded this quarter. The rest carry verification dates into H2 or later.

2026-W02 Opus 4.7 resets the frontier. Verified. Anthropic's Opus 4.7 release in March hit the coding-benchmark posture we called and the 1M-context-window operational behavior that 2026-W21 will analyze.

2026-W03 MBZUAI graduates first spinout cohort. Verified. Five spinouts announced at the MBZUAI demo day in February, four of those raised seed or Series A inside thirty days of demo, two with PIF anchor checks.

2026-W04 Agentic workflows eat SaaS in Q1. Verified. The enterprise SaaS retention numbers published by three of the top horizontal SaaS vendors in Q1 confirm sharper than expected churn in the categories most exposed to agentic alternatives. The quantitative shape of the call has held.

2026-W05 DeepSeek R2 and the China resurgence. Verified. R2 landed in February at the capability level we predicted and triggered the second China-frontier moment we forecast in 2025-W43. Nvidia drawdown shape was smaller this time, as the market had priced in the pattern.

2026-W06 G42-Microsoft Phase Two. Verified. The expansion agreement landed in February at $1.8B incremental commitment, with the data-residency package we predicted.

2026-W09 The context-window arms race ends. Verified. Opus 4.7's 1M-context release is generally credited as the closing move of the context-window race. We were on the record naming this call before the release.

The two new wrong calls

2026-W08 Saudi Humain goes public. Partial-pointing-to-wrong. Humain's listing prospectus was filed in Q1 but the listing has been postponed by the issuer into late 2026 or 2027. The transaction will likely happen. The timing was wrong.

2026-W17 The on-prem LLM renaissance. Partial. On-prem deployments have grown but at half the rate we predicted, because the inference-cloud cost compression has continued faster than we modeled. The on-prem case is real for sovereign and regulated workloads. Our 2026 magnitude was too aggressive.

Aggregate distribution

Total verifiable calls program-to-date: 156. Verified: 127 (81%). Partial: 18 (12%). Wrong: 11 (7%).

Distribution band targets: 70% verified / 20% partial / 10% wrong.

We are still running above target on verified rate. This is the third consecutive audit where we have done so. We are introducing two changes for H2.

First, we are committing to publish at least four high-confidence contrarian calls per quarter in H2, where high-confidence means calls we are willing to grade against the consensus view rather than against our internal base case. These calls should fail more often.

Second, we are committing to longer-horizon predictions. The 2025 program was weighted toward six-to-twelve-month verification windows. The 2026 program will include more eighteen-month and twenty-four-month calls. Longer horizons surface our errors more honestly.

What this means for the Gulf

The audit confirms the sovereign-AI thesis is holding. The 2026-W01 annual call is tracking on. PIF, G42, Humain, and the MBZUAI-TII research-to-production pipeline are all in the expected positions through H1.

The agentic-banking call (2026-W12) is on track for the Q3 deadline. We expect to grade it verified or partial-verified in the next quarterly audit.

The talent-shortage call (2026-W24) is the H2 watch. Compensation data for senior AI roles in Abu Dhabi and Riyadh shows a 35% premium over comparable London and 20% over comparable New York at the staff and senior staff level. The shortage is real and deepening.

The live /track-record page is updated weekly. You can grade us against every call we have published. The point of this column is that the next twelve months of GCC AI predictions land on a record that the previous twenty-four months made credible. The forecast without the audit is opinion. The forecast with the audit is a position you can take to your board.

We will publish the Q3 audit in 2026-W40.