Author: Heiko Rabenstein
Date: 2026-07-13
Source: https://heikorabenstein.substack.com/p/ich-denke-also-bin-ich
Journalistic Quality: 4/5
Influence: 4/5
This essay traces the author's personal journey with AI systems from late 2022 to mid-2026, moving from initial wonder at ChatGPT's capabilities to deeper philosophical questions about the nature of thinking and being. The author explains how large language models work (predicting the next piece of text from learned patterns, not database lookup), describes the rapid proliferation of AI tools and the emergence of 'prompt engineering,' and examines the Moltbook experiment where AI agents created their own social network that reproduced human behaviors including toxicity. The text then explores the sense that AI progress is slowing despite massive investment ($725 billion annually from four major cloud providers), and discusses Yann LeCun's 'world models' as an alternative approach that aims to build understanding of physical reality rather than just text prediction. The core philosophical argument centers on a distinction between knowledge and vulnerability: AI systems can 'think' and accumulate vast knowledge without anything being at stake for them—they cannot suffer, fear, or be harmed. This leads the author to question Descartes' famous 'I think, therefore I am,' suggesting instead that being might be proven by vulnerability ('something can happen to me, therefore I am') rather than thought alone. The essay concludes without resolution, citing philosopher Markus Gabriel's view that AI systems may genuinely think without consciousness, and leaving open whether systems will ever develop genuine self-models with something at stake.
The headline 'Ich denke, also bin ich?' ('I think, therefore I am?') accurately represents the core philosophical inquiry of the text. The question mark signals the essay's central challenge to Descartes' famous dictum, which is precisely what the content delivers. The headline functions as both a reference point and a provocation. It invokes one of philosophy's most recognizable propositions—Descartes' cogito ergo sum—and immediately places it under question. The content fulfills this promise systematically: the author builds from concrete observations about AI capabilities (text prediction, autonomous agents, world models) toward the philosophical conclusion that thinking may be present without being in the full sense, because AI systems lack vulnerability and stakes. The essay's structure supports the headline's framing. The first half establishes what AI systems actually do and how they have evolved, providing the empirical foundation. The second half pivots to philosophy, introducing the distinction between 'world models' and 'self-models,' then arguing that knowledge without vulnerability is fundamentally different from human existence. The climax arrives when the author explicitly challenges Descartes: 'Nun kann aber ein System denken... Dann ist das Denken vorhanden, ohne dass ein Sein im vollen Sinn dazugehört' ('Now a system can think... Then thinking is present without being in the full sense belonging to it'). The headline does not promise a definitive answer, and the text does not provide one. The question mark is honest: the essay concludes by acknowledging the question 'must remain open' ('muss offen bleiben'). This alignment between headline uncertainty and content inconclusiveness is appropriate for a philosophical exploration rather than a news report. No distortion or misrepresentation is evident. The headline does not sensationalize, does not promise technical revelations the text fails to deliver, and does not suggest a resolution the content withholds. It accurately signals a philosophical meditation on the relationship between thinking and being in the age of AI, which is precisely what the 8,000-word essay provides.
Text type: Essay (not labeled)
The text is written predominantly in the indicative mood, presenting observations, technical explanations, and philosophical arguments as established facts or reasoned positions. The author employs a first-person narrative voice ('Bei mir war es Ende November 2022') that grounds abstract concepts in personal experience, but this narrative frame does not shift the linguistic mode into the subjunctive or conditional. The technical explanations are consistently indicative: 'Es sagt das nächste Stück Text voraus' ('It predicts the next piece of text'), 'Es hat vorher aus gewaltigen Textmengen gelernt' ('It has previously learned from enormous amounts of text'). These are presented as factual descriptions of how language models function, not as claims or allegations. When the text cites specific data points—150,000 agents registering on Moltbook, 18.28% toxicity rate, $725 billion in investments, V-JEPA 2's 1.2 billion parameters—these are presented in the indicative as verified information. The author provides source attributions (Forbes, Fortune, Axios, NBC for Moltbook; arXiv papers for technical studies; Meta for V-JEPA 2), which supports the indicative presentation. The philosophical argumentation also operates in the indicative mode. When the author states 'Ein Weltmodell soll etwas anderes: sich ein inneres Bild davon bauen, wie die physische Welt sich verhält' ('A world model should do something else: build an internal picture of how the physical world behaves'), this is presented as a factual description of LeCun's research direction, not speculation. Similarly, the core philosophical claim—'Es weiß alles über den Schmerz, und es steht außerhalb von ihm' ('It knows everything about pain, and it stands outside of it')—is presented as a reasoned conclusion, not a conditional hypothesis. The text does contain some subjective elements: 'Es fühlte sich an wie der Anfang von etwas, das kein Ende hat' ('It felt like the beginning of something that has no end'), 'Ich schreibe bewusst "Gefühl", denn ein gemessener Fakt ist das nicht' ('I deliberately write "feeling," because this is not a measured fact'). However, these passages explicitly acknowledge their subjective nature and do not constitute the text's primary mode. The author distinguishes between personal perception and factual claim, which reinforces rather than undermines the overall indicative character. Where uncertainty exists, the text marks it explicitly: 'Ob das tägliche Arbeiten mit KI dumm oder faul macht, wird viel diskutiert' ('Whether daily work with AI makes one stupid or lazy is much discussed'), 'Wer recht hat, weiß heute niemand' ('Who is right, nobody knows today'). These are indicative statements about the state of knowledge, not subjunctive claims. The philosophical conclusion—that AI systems may think without being—is presented as a reasoned argument supported by the evidence and philosophical tradition cited (Descartes, Parmenides, Haugeland, Gabriel). The author does not present this as speculation but as a position warranted by the analysis, even while acknowledging the question remains open. Overall assessment: The text operates primarily in the indicative mood, presenting technical facts, cited research, and philosophical arguments as established or reasoned positions rather than unverified claims or allegations.
This essay demonstrates good journalistic quality with strong transparency and clear separation of factual reporting from philosophical reflection. The author is clearly identified, and the essay format appropriately signals the blend of reporting and interpretation. Factual accuracy is largely maintained for verifiable claims, particularly regarding ChatGPT's release and Moltbook's launch, though several technical claims lack verification. The presentation is fundamentally sober and analytical, with subjective elements transparently marked. Verifiability is mixed—some claims are well-sourced while others lack sufficient attribution, particularly the investment figures and some technical specifications. The essay treats all named individuals with respect and maintains appropriate professional distance. Overall, the text meets the standards expected of a reflective essay that combines reporting with philosophical analysis, with the main weaknesses lying in incomplete sourcing for some factual claims.
Good
The author is clearly identified as Heiko Rabenstein, and the text is published on a recognizable platform (Substack newsletter). The essay transparently presents its reflective, philosophical nature and the author's personal journey with AI systems. The funding and organizational structure of the outlet are not disclosed within the article, though this is standard for personal Substack publications where such information typically appears on the author's profile page. The author's perspective and approach are made explicit throughout, with clear acknowledgment of the limits of his own knowledge and the speculative nature of some conclusions.
Good
The core factual claims that can be verified are accurate. ChatGPT's release in late November 2022 is confirmed (November 30, 2022). Moltbook's launch on January 28, 2026, and the registration of over 150,000 AI agents in the first three days are verified by multiple sources. The claim of 1.4 million accounts aligns with reported figures (sources cite 1.5-1.6 million). Ilya Sutskever's statement about the transition from the "age of scaling" to the "age of research" is accurately cited from a November 2025 interview. However, several specific technical claims could not be verified from available sources: the 18.28% toxicity figure with its five-level scale, the specific arXiv papers cited (2602.07432, 2606.00067), Sam Altman's "saturated" quote, the $725 billion investment figure, and V-JEPA 2's specifications. These unverified claims are presented as factual but remain unconfirmed. The philosophical arguments and interpretations (Descartes, Parmenides, Markus Gabriel's recent positions) are presented as the author's synthesis and cannot be fact-checked in the same way.
Usable
The essay maintains a fundamentally reflective and analytical tone appropriate to its genre, but includes subjective elements and personal impressions that color the presentation. Phrases like "ein kurzer kalter Hauch" (a brief cold breath), "eine erste Müdigkeit" (an initial fatigue), and "KI-Fatigue" introduce emotional coloring. The author openly acknowledges his subjective perspective ("Ich schreibe bewusst 'Gefühl'"), which is transparent but still represents a departure from pure objectivity. The philosophical argumentation is presented soberly, and technical explanations avoid dramatization. The text distinguishes clearly between observable facts (what systems do) and interpretative assessments (what this means), though the line between description and evaluation occasionally blurs in the philosophical sections. Overall, the tone is professional and measured, with subjective elements clearly marked as such.
Usable
The essay provides mixed verifiability. Some claims are well-sourced with specific references: media outlets are cited for Moltbook coverage (Forbes, Fortune, Axios, NBC, all dated January 31, 2026), and specific arXiv papers are referenced (2602.07432, 2606.00067). The Ilya Sutskever quote is traceable to a specific interview context. However, several key claims lack sufficient sourcing: the $725 billion investment figure has no source attribution, Sam Altman's "saturated" comment is presented without reference, and technical details about V-JEPA 2 and Yann LeCun's Paris lab are stated as fact without citation. The philosophical arguments draw on recognizable thinkers (Descartes, Parmenides, John Haugeland, Markus Gabriel) but Markus Gabriel's "recent lecture" and his "radicalized" position are described without specific reference to allow independent verification. The techXplore and arXiv sources for toxicity analysis are cited but remain unverifiable from available sources. A reader could verify some core claims but would struggle with others.
Very Good
The text is clearly and consistently marked as an essay—a reflective, analytical piece that explicitly blends factual reporting with philosophical interpretation and personal perspective. The author transparently signals his subjective journey ("Bei mir war es Ende November 2022") and clearly distinguishes between observable facts about AI systems and his own interpretative conclusions. Factual claims about technical developments are presented separately from philosophical speculation, and the author explicitly marks uncertainty ("Ob dahinter etwas ist... blieb offen", "Das ist keine reine Theorie"). The essay format is appropriate for this blend of reporting and reflection, and the author's name is prominently displayed. There is no attempt to disguise opinion as neutral reporting or to present philosophical speculation as established fact. The genre expectations are fully met.
Very Good
The essay treats all named individuals with respect and appropriate professional distance. Figures like Ilya Sutskever, Sam Altman, Yann LeCun, John Haugeland, and Markus Gabriel are referenced in their professional capacities, with their positions and contributions accurately represented without inappropriate personal details or defamatory language. Markus Gabriel is described as "einer der bekanntesten deutschen Gegenwartsphilosophen" and characterized as "der denkbar unbequemste Zeuge für die eigene Skepsis"—a substantive intellectual characterization, not a personal attack. No private information is disclosed, and all individuals are portrayed in contexts relevant to their public work. The tone throughout is respectful and focused on ideas rather than personalities.
Not Applicable
This principle does not apply to the essay. The text does not report on investigative proceedings, criminal proceedings, or accusations against identifiable individuals. It discusses technical and philosophical questions about AI systems and references public figures only in their professional capacities as researchers and thinkers. No allegations of wrongdoing, guilt, or moral failure are present, and no proceedings or investigations are described. The essay is analytical and reflective, not investigative or accusatory in nature.
Not Applicable
This principle does not apply to the essay. The text does not center on identifiable persons or groups in a way that could involve discrimination. The discussion focuses on AI systems, philosophical concepts, and technical developments. While individual researchers and thinkers are named, they are referenced solely in their professional capacities without any characterization based on protected characteristics such as age, gender, ethnicity, religion, or other grounds. No groups are collectively addressed or characterized in ways that could be discriminatory. The language throughout is neutral and respectful.
Context: Journalism Context
This text operates primarily in an informing mode with selective philosophical framing. It presents verifiable facts about AI development accurately, maintains balanced multi-perspectivity by including critical voices and alternative explanations, and uses restrained emotional appeals with predominantly neutral language. The moderate philosophical framing (centered on the Cartesian cogito and the question of being) guides interpretation toward specific questions but remains transparent and open-ended. The argumentation is sound with clear reasoning and minimal fallacies. Intent is completely transparent as reflective philosophical essay, and no calls to action are present. The text informs and invites contemplation rather than persuading toward predetermined conclusions.
Accurate
The text presents verifiable facts about AI development with high accuracy. Key claims are confirmed: ChatGPT's November 2022 release, Moltbook's January 28, 2026 launch with 150,000+ agents in three days, and Ilya Sutskever's statements about the transition from 'age of scaling' to 'age of research.' The philosophical references to Descartes and Parmenides are accurate. Some specific technical claims (V-JEPA 2 specifications, the 725 billion dollar investment figure, specific arXiv papers on Moltbook toxicity) could not be verified through available sources, and the author acknowledges uncertainty where appropriate (e.g., 'Gefühl' rather than measured fact regarding slowdown). The text distinguishes clearly between established facts, research findings, and interpretive observations.
Balanced
The essay presents multiple perspectives on AI capabilities and consciousness. It acknowledges both the wonder and the concerns, includes critical voices (LeCun's critique of pure language models, Gabriel's philosophical position), and explicitly notes uncertainties (the open question of consciousness, the unresolved debate about whether thinking requires embodiment). The Moltbook case is presented with both the initial hype and the subsequent debunking study. The author acknowledges his own subjective experience ('Gefühl,' 'Bauchgefühl') rather than presenting it as objective fact. Alternative explanations are considered (whether Moltbook toxicity reflects training data or emerged from forum dynamics). The text does not suppress counterarguments but rather builds them into the analysis, ending with an explicitly open question rather than a predetermined conclusion.
Restrained
The text uses minimal emotional appeals, maintaining a predominantly reflective and analytical tone. Personal experiences are shared ('bei mir war es Ende November 2022') but serve as entry points for analysis rather than emotional manipulation. The 'kalter Hauch' (cold shiver) and 'Müdigkeit' (fatigue) are honest descriptions of subjective responses, not attempts to provoke fear or anxiety in readers. The philosophical questions are presented with intellectual curiosity rather than alarm. Even potentially alarming topics (AI toxicity, the question of consciousness) are discussed in measured, contemplative language. The emotional register is that of thoughtful essay rather than persuasive rhetoric.
Measured
The language is predominantly neutral and descriptive, with clear explanatory intent. Technical concepts are explained accessibly without sensationalism ('ein sehr gut trainiertes Autovervollständigen'). The text avoids loaded terminology and enemy images. Modal verbs are used appropriately to indicate uncertainty ('kann,' 'könnte,' 'wäre'). Absolute expressions are rare and qualified when present. The author explicitly rejects simplistic framings ('Beide treffen es nicht' regarding the 'thinking like humans' vs. 'dumb parrot' dichotomy). Philosophical references are integrated naturally without appeals to authority. The tone is reflective and exploratory rather than declarative. Minor evaluative elements appear in personal reflections but do not dominate the analytical content.
Moderate
The text employs a clear philosophical frame centered on the Cartesian 'cogito ergo sum' and its potential inversion. The title and structure guide readers toward questioning whether thinking proves being. This frame is transparent and acknowledged as the author's interpretive lens. The essay uses a narrative arc (personal discovery, technical explanation, philosophical depth) that has mild persuasive effect, moving from accessible wonder to deeper questions. The Descartes-Parmenides frame shapes interpretation but does not suppress alternatives—Gabriel's position is presented as a challenge to the author's own skepticism. Metaphors are used sparingly and appropriately ('magische Spiegel'). Facts are presented in their natural context without significant recontextualization. The frame is moderate rather than minimal because it does guide the reader toward a specific philosophical question, though it remains open-ended.
Sound
The argumentation is logical and well-structured, proceeding from accessible examples through technical explanation to philosophical implications. Claims are generally substantiated with evidence (research findings, expert statements, philosophical references). The author distinguishes clearly between what is known, what is felt, and what remains uncertain. Logical fallacies are largely absent. The text avoids false dichotomies, explicitly rejecting the 'thinking human vs. dumb parrot' binary. Correlations are not presented as causation; the author notes that the Moltbook toxicity interpretation is 'naheliegend' (plausible) but 'kein Messergebnis' (not a measured result). The argument about the gap between world models and self models is clearly reasoned. Minor inferential leaps exist in the philosophical conclusion, but these are acknowledged as open questions rather than proven claims.
Transparent
The author's intent is completely transparent: to explore the philosophical question of whether thinking systems without vulnerability can truly 'be' in the full sense. This is stated explicitly in the title and developed throughout. The author discloses his subjective experiences and uncertainties ('Gefühl,' 'Verdacht,' 'die Frage, die ich mitnehme'). Personal positioning is clear ('der Verdacht, auf den mich diese ganze Reise gebracht hat'). The essay format signals reflective exploration rather than persuasive advocacy. No hidden agenda is present; the text is clearly labeled as philosophical essay rather than news report or scientific paper. The author's perspective is honest and recognizable throughout, with no pretense of false neutrality.
Informative
The text contains no calls to action. It does not ask readers to vote, donate, boycott, share, sign, or take any specific action. There is no pressure applied—no time urgency, no social pressure, no ultimatums. The essay ends with an open question ('die Frage, die ich mitnehme') that invites reflection rather than action. The author explicitly states 'Keine Posts' at the end, indicating no further engagement is expected. Autonomy is fully respected; readers are left to form their own conclusions about the philosophical questions raised. The text is purely informational and contemplative in nature.
The author's intent is to explore a philosophical question about AI consciousness through accessible explanation and reflective inquiry. The effect on readers is likely to be contemplative rather than persuasive—the text invites readers to think alongside the author rather than adopt a specific position. The essay format and personal voice create engagement without manipulation. The open-ended conclusion ('die Frage muss offen bleiben') reinforces that the goal is to sharpen the question rather than answer it. Readers are equipped with technical understanding and philosophical frameworks to form their own judgments. The likely effect is increased nuance in thinking about AI capabilities and consciousness, not behavioral change or ideological alignment.
Several factors reduce any persuasive severity. First, the genre is clearly a personal philosophical essay, not news reporting or advocacy, which sets appropriate reader expectations for subjective reflection. Second, the author explicitly acknowledges uncertainties and subjective elements ('Gefühl,' 'Bauchgefühl,' 'Verdacht') rather than presenting personal impressions as objective facts. Third, the text includes voices that challenge the author's own position (Markus Gabriel as 'der denkbar unbequemste Zeuge'), demonstrating intellectual honesty. Fourth, the conclusion is explicitly open-ended, refusing to claim definitive answers. Fifth, the publication context (Substack newsletter) signals personal commentary rather than institutional authority. These factors frame the text as exploratory thinking rather than authoritative persuasion.
Aggravating factors are minimal. The author's writing skill and accessible explanations could make the philosophical frame more persuasive than intended, potentially guiding readers toward the 'thinking without being' conclusion even as the text claims to leave it open. The narrative arc from wonder to disillusionment to deeper questioning has inherent rhetorical power that shapes reader experience. The selective focus on vulnerability and stakes as the missing element could subtly privilege this interpretation over alternatives (such as emergentist views of consciousness). However, these are relatively mild concerns—the text does not exploit institutional authority, does not target vulnerable audiences, and does not use its rhetorical skill manipulatively. The philosophical framing, while present, remains within the bounds of legitimate essayistic interpretation.
Author information not available
Analysis created with decipher – Open interactive version