Get Reality
Updated 30 July 2026
Next update 6 August 2026

The Trust Stack · Weekly Signal Board

Proving it is the product.

AI capability is no longer scarce. What is scarce is the ability to prove what an AI did, why it did it, and that the resources behind it were legitimate.

$1T

Market capitalisation erased from AI chip and memory names in the last week of July 2026 - while several of those companies posted record profits. The market has begun pricing compute as a commodity whose returns must be demonstrated rather than assumed.

Evidence

The numbers that survive verification.

Every figure below was traced to a primary source this week. Anything that could not be traced was removed rather than softened - see the integrity note at the foot of this page.

GenAI deployments carrying LLM observability investment, by 2028

50%

Up from 15% as of March 2026. Explainability is the stated driver.
Gartner, 30 Mar 2026 · projection

Enterprise applications embedding task-specific AI agents, by end of 2026

40%

Up from under 5% in 2025. No independent mid-year measurement exists.
Gartner, 26 Aug 2025 · projection

Cost of human-delivered work that AI can begin to absorb

$4.6T

Salaries across four job families plus outsourced IT and business-process spend. Not an AI revenue forecast.
Foundation Capital, Apr 2024 · constructed TAM

Enterprise SaaS spend shifting to usage, agent or outcome pricing, by 2030

40%

Stated as a floor - "at least" 40%.
Gartner via Deloitte TMT Predictions 2026 · projection

Big-four hyperscaler AI infrastructure spend in calendar 2026

$650B

Alphabet, Amazon, Meta, Microsoft. Company guidance since has run to $660-700B.
Bridgewater via Reuters, 23 Feb 2026 · projection

Tracked AI-crypto projects already dead

70%

Measured, not forecast. Average survival 140 days across 338 tracked projects. Selection bias applies.
Ventureburn tracking, upd. 6 Jul 2026 · measured

The stack

Five layers, and where the evidence is moving.

The framework is stable. What changes week to week is the strength of the evidence behind each layer. Direction reflects developments in the trailing quarter, not price action.

Evidence direction is marked with a symbol and a word as well as a colour, so it never depends on colour alone.
LayerWhat lives hereEvidence
1 · Compute Physical and distributed infrastructure doing the work Weakening
2 · Verification Cryptographic proof that the work happened as claimed Strengthening
3 · Protocol How agents reach tools and each other - MCP, A2A Strengthening
4 · Application Domain-deep agents that do the actual job Mixed
5 · Edge & behaviour Where AI meets the physical world and real behaviour Mixed

What moved

This week, and the quarter behind it.

  • 27 Jul 2026

    The EU just deferred the thesis's biggest tailwind by sixteen months

    Against

    Regulation (EU) 2026/1744 entered into force, moving the AI Act's high-risk obligations from 2 August 2026 to 2 December 2027, and embedded-product obligations to August 2028. Logging and traceability duties were postponed, not weakened - Article 12 still mandates lifetime event logging, with penalties to 3% of worldwide turnover. Article 50 transparency still starts 2 August 2026. Any business case built on an August 2026 compliance deadline needs rebuilding.

    Official Journal, 24 Jul 2026; in force 27 Jul 2026

  • 24-29 Jul 2026

    Over $1 trillion comes off AI chip names despite record profits

    For

    Nvidia, SK Hynix, Samsung, Micron, AMD, ASML, Arm and SoftBank all fell hard, with a record wave of trading halts in Korea. Infrastructure announcements no longer automatically lift semiconductor shares while financing costs, utilisation and eventual revenue remain uncertain. Read carefully, this is a repricing of expectations rather than collapsed demand - but it is the market demanding that compute demonstrate its returns.

    CNBC, 29 Jul 2026

  • 23 Jul 2026

    The IETF votes to standardise agent communication - but rejects the scope

    For

    At IETF 126 in Vienna, a working-group-forming session on agent communication protocols recorded 154 in favour of chartering against 51 opposed - then rejected the proposed initial scope 124 to 38. Strong consensus that this must be standardised; no agreement yet on what. The named gaps are cross-domain identity federation, multi-hop lifecycle management, user confirmation before irreversible actions, and protocol-level attribution.

    IETF 126 agentproto minutes, 23 Jul 2026

  • 28 Apr 2026

    FIDO, Visa, Mastercard and Google start building the verification primitive

    For

    FIDO chartered an Agentic Authentication working group with Google, Mastercard, Visa, CVS Health, OpenAI, Amazon and Okta in leadership. Its three workstreams - verifiable user instructions, agent authentication, trusted delegation for commerce - are the thesis, written by a standards body. Mastercard's Verifiable Intent binds identity, instruction and outcome into one tamper-resistant record. The caveat cuts both ways: standards bodies making verification free is validation and margin pressure at once.

    FIDO Alliance, 28 Apr 2026

  • 22 Apr 2026

    Google folds agent identity and audit logging into the platform

    Against

    The Gemini Enterprise Agent Platform ships cryptographically signed agent cards, zero-trust architecture across decentralised agent systems, and IAM-integrated audit logging as native features. This is the strongest argument against an independently investable trust layer, and it deserves a direct answer: what survives is verification a platform cannot credibly perform on itself. A hyperscaler attesting to its own compute is a self-certification. Regulated and adversarial buyers will not accept one.

    Google Cloud Next, 22 Apr 2026

What this page does not show

The framework is public. The screen is not.

This board tells you what I look for and why, and shows the evidence I check it against. It deliberately stops short of the part that does the actual work:

  • The green flags that earn a second meeting, and the disqualifying signals that end one
  • The stack checklist, scored and weighted by layer position
  • The diligence question set - including the one question that now separates a durable verification business from a feature a hyperscaler ships for free
  • How any of it applies to a specific company

That work is what I am engaged for, in cash or in equity. If you are building in this stack, or writing cheques into it, the useful conversation starts with the thing you are actually worried about.

Get Real Now

Integrity note

What I removed from my own thesis this week.

The April edition of this thesis carried twenty-two market claims. On re-verification, seven were dropped and eight were restated. One figure was overstated by roughly 170 times - token trading volume had been read as protocol revenue. One capability was described as shipped that has never shipped. One regulatory timeline had been rewritten by law three days before this update. Every failure shared a cause: the number traced through a blog rather than to a primary source, and mutated at each hop.

I publish that because a firm called Get Reality does not get to apply the standard only to other people's decks. The full correction log lives in the private edition, and any figure on this page that I cannot re-trace next week will come off it.