lifearchitect.ai

What is LifeArchitect.ai?

LifeArchitect.ai is the personal research hub of Dr Alan D. Thompson, an Australian AI researcher whose work focuses on tracking and explaining the capabilities of post-2020 AI systems, up to and beyond AGI and superintelligence. The site collects his papers, articles, videos, model comparisons and regular "Memo" briefings in one place, and its stated audience includes major AI labs, governments, research institutes and general readers trying to follow the field.

H3 What you'll actually find there

• Model tracking: A large models table (described as covering 1,000+ models and tens of thousands of data points), plus timelines, dataset tables, an AGI countdown and an ASI checklist.
• Explainers and papers: Titles like "What's in GPT-5?", "What's in Grok?", "What's in my AI?" and "The psychology of modern LLMs" dig into what is inside specific systems.
• Recurring reports: Twice-yearly AI reports (mid-year and end-of-year editions going back several years) and The Memo, a regular newsletter-style briefing.
• Media and commentary: Video content, media releases and essays, including more speculative pieces on superintelligence and its implications.

H3 Who it suits, and the trade-off It is most useful if you want a single researcher's synthesised, frequently updated read on where AI stands, especially if you prefer structured tables and timelines over scattered news. The trade-off is that this is largely one person's curated perspective, with an openly pro-acceleration, "get ready for superintelligence" tone — valuable as a lens, but worth pairing with primary sources and more sceptical coverage.

H3 Next step If you want a quick sense of the site's style, start with a recent AI report or a "What's in [model]?" paper, then check the models table for the comparison view. For independent technical detail on specific models, cross-reference official documentation from the labs themselves, such as OpenAI or Google DeepMind, and for broader industry news see Ars Technica.

Who is Dr Alan D. Thompson and why is he cited by Microsoft and major AI labs?

Dr Alan D. Thompson is an Australian AI researcher and writer who runs LifeArchitect.ai, a long-running research and publishing project focused on large language models and the trajectory toward AGI and superintelligence. His reputation rests on volume, consistency and timing: the site describes more than 100 papers and articles, 300 videos and regular editions of a briefing called The Memo, with model-by-model analyses covering systems such as GPT-3.5, GPT-4, GPT-5, Gemini, Grok, Chinchilla, ERNIE and o-series models.

Why labs and institutions cite him

The value is in synthesis rather than primary research. His site maintains reference resources that are tedious to assemble yourself — a models table described as covering 1,000+ models and 40,000+ LLM data points, a timeline, a datasets table, an AGI countdown, an ASI indicator checklist and periodic AI reports going back to mid-2021. Papers such as "What's in my AI?", "What's in GPT-5?" and "What's in Grok?" attempt to consolidate what is publicly known about a given model in one place. For a Microsoft researcher, government analyst or lab communications team, that is a convenient citation for context, chronology or public-facing framing, even if it is not a substitute for a technical paper.

How to use the site

• If you need a quick orientation on a new model release, start with the model-specific paper or the models table.
• If you are tracking the broader debate, the dated AI reports and media releases (for example the 2021 "AI fire alarm" and 2023 "leaders guilty of negligence" statements) show how his position has evolved.
• If you want his own forecast, the mid-2026 retrospective and the AGI/ASI checklists are the most direct expressions of it.

A practical caveat

Thompson is an advocate as well as a chronicler. The site's framing is openly accelerationist — his July 2025 quote urges readers to "be ready" for superintelligence rather than to look for faults in it. That makes his work useful for mapping the optimistic case and for consolidated model data, but you should pair it with independent benchmarks and peer-reviewed evaluations before treating any claim as settled. A reader deciding whether to subscribe should ask whether they want a curated, opinionated overview or raw evaluation data; the site is strongest for the former.

How can I use LifeArchitect.ai's models table and timeline to track AI progress like GPT-5 and AGI?

Use LifeArchitect.ai's Models table as a structured reference for comparing named systems, and its Timeline and countdown sections as a way to place those systems in sequence rather than treating each release as an isolated event. The page evidence describes a Models table covering 1,000+ models and 40,000+ LLM data points, plus a Timeline, a Countdown to AGI, and an ASI checklist — so the practical workflow is: look up a model, then read the timeline around it to see what changed and what was expected next.

H3 How to use each part

• Models table: Best for questions like "what exists, from which lab, and how does it compare to GPT-4 or Gemini?" Use it as a lookup layer, not a narrative.
• Timeline: Best for sequencing — seeing that GPT-3.5, GPT-4, GPT-5 and GPT-6 are treated as points on a line, alongside Chinchilla, ERNIE, Gemini, Grok, Nova and the o-series.
• Countdown to AGI / ASI checklist: Best for tracking claims about thresholds. Treat these as one analyst's framing of progress, not as settled definitions.
• Papers and reports: The site lists model-specific papers (for example "What's in GPT-5?" and "What's in Grok?") and periodic AI reports from mid-2022 through end-2025, which let you compare what was predicted with what later appeared.

H3 A concrete reader scenario Suppose you want to understand where GPT-5 sits. Start with the Models table entry, then open the associated paper, then move to the Timeline to see what preceded and followed it. Finally, check the AGI countdown and ASI checklist to see how that release is positioned against broader capability claims. This gives you three layers — specification, chronology, and interpretation — instead of a single headline.

H3 Trade-offs and decision criteria

• If you need quick orientation, the table and timeline are the fastest entry points.
• If you need depth on one system, the model-specific papers are more useful than the table alone.
• If you need neutral consensus, look elsewhere as well: the site is explicitly positioned around one author's analysis, and its AGI/ASI framing reflects that viewpoint.
• If you need regular updates, the recurring Memo and periodic reports are the relevant format.

For cross-checking, the official sources of the labs themselves are worth pairing with this: OpenAI, Google DeepMind, and Anthropic. For a broader research index, arXiv is a reasonable complement.

What do LifeArchitect.ai's AI reports and The Memo subscription cover that I can't get from mainstream AI news?

LifeArchitect.ai is built for depth and forward tracking rather than daily news. Its AI reports and The Memo subscription are aimed at readers who want structured, longitudinal analysis of where AI capabilities are heading—not just what shipped this week.

What the reports and The Memo cover

• Periodic AI reports (mid-year and end-of-year editions going back to 2021) that consolidate model progress, benchmarks and capability shifts into a single retrospective. Mainstream outlets rarely publish this kind of recurring, comparable snapshot.
• Model-level breakdowns such as "What's in GPT-5?", "What's in Grok?" and "What's in my AI?"—papers that dissect what is actually inside a given system rather than reporting its launch.
• Tracking tools including a models table (the site claims 1,000+ models and 40,000+ LLM data points), a timeline, datasets table, a countdown to AGI, and an ASI checklist.
• The Memo, a regular subscription edition that packages these insights for readers following the field closely.

How that differs from mainstream AI news

Who it suits

If you are a researcher, analyst, policy staffer or technically literate enthusiast who needs a consistent reference point across model generations, the recurring reports and model tables are the main draw. If you just want to know what happened today, mainstream coverage will be faster and cheaper.

A practical next step

Pick one recent report edition and one model paper (for example the GPT-5 or Grok breakdown) and read them together. If the combination of a periodic snapshot plus a deep model teardown is more useful to you than a news feed, the subscription is likely worth evaluating; if not, the free summaries and tables may be enough.

How does LifeArchitect.ai's ASI checklist and countdown to AGI help me prepare for superintelligence?

LifeArchitect.ai's ASI checklist and countdown to AGI are preparation tools rather than predictions you can set a clock by. The countdown frames how close the site's author, Dr Alan D. Thompson, judges artificial general intelligence to be, while the ASI checklist breaks "superintelligence" into observable indicators you can track over time. Used together, they turn a vague worry about the future into a set of things to watch and act on.

LifeArchitect.ai

What each tool gives you

• Countdown to AGI: a directional estimate of how near AGI is, useful for deciding how urgently to act rather than for precise scheduling.
• ASI checklist: a list of indicators (the page references "Alan's ASI indicators (first 50)") that describe what superintelligent capability might look like as it emerges.
• Supporting tables: the Models Table (1,000+ models and 40,000+ LLM data points), Timeline, and Datasets Table let you check whether the checklist items are actually showing up in real systems.

The practical value is the pairing: the countdown tells you when to pay attention, the checklist tells you what to look for. Neither is a forecast you should treat as settled.

How to actually use them

1. Skim the checklist and mark the indicators that would affect your work, field or family first.
2. Check the Models Table and Timeline every few months to see whether those specific indicators are moving.
3. Read the regular The Memo editions or the periodic AI reports (the page lists Mid-2025, End-2024 and earlier) to get the author's updated read rather than relying on one snapshot.
4. Turn each relevant indicator into one concrete preparation step — a skill to build, a tool to trial, a decision to revisit.

Who benefits most

The main caveat: this is one expert's framework, published alongside his own papers and media releases, and it leans toward an optimistic, fast-timeline view. Treat the checklist as a lens for organising your own thinking, not as a neutral consensus.

A useful next step: pick three checklist indicators tied to your own work, note today's state of each, and set a reminder to re-check them against the Models Table in six months. That converts the countdown from something you read into something you act on.

What are the key findings in LifeArchitect.ai's papers on Grok, GPT-5, and the psychology of modern LLMs?

LifeArchitect.ai's papers on Grok, GPT-5 and the psychology of modern LLMs are best understood as tracking work rather than one-off research papers. The site's own framing is that it maintains an ongoing record of what goes into each major model, alongside a large models table and a running AI report series, so the findings accumulate across releases instead of arriving as a single conclusion.

What each strand covers

• Grok — The page lists "What's in Grok?" and an independent Grok 5 paper. The focus is on what the model is built from and what has changed between versions, rather than benchmark scores alone.
• GPT-5 — "What's in GPT-5?" sits alongside GPT-3.5, GPT-4 and GPT-6 in the models list. The intent is a plain account of the model's composition and positioning, written for readers who need to brief others.
• Psychology of modern LLMs — This is the conceptual piece: how these systems behave, rather than what they score. It pairs with the site's alignment, "The Gap" and AI + IQ material.

Why the framing matters

Alan D. Thompson's stated position is that capability is arriving faster than rollout, and that fault-finding with superintelligence is a poor use of time because the faults get solved quickly. That view shapes the papers: they lean toward capability tracking and preparation rather than critique. Read them as an informed insider's running log, not as peer-reviewed studies.

How to use it

If you need a quick orientation before a strategy or policy discussion, start with the most recent AI report, then read the model-specific paper for whichever system you actually use. If you want the deeper argument, go to the psychology and alignment pieces. For comparison against other trackers, Epoch AI publishes model and compute trend data, and arXiv is where the underlying technical papers live.

A caveat worth holding

The site is written by one author with a strong point of view, and its reach into major labs and government bodies is part of its own description. That makes it fast and well-connected, but it is not a neutral scoreboard — cross-check specific claims against primary sources before citing them.