Labs · Feb 27, 2025
Yoav argues in AI21 Labs’ excerpt for building complete AI systems
Markdown · JSON · Original source
Published February 27, 2025, this excerpt from Yoav’s contribution to Google Cloud’s “Future of AI: Perspectives for Startups 2025” report argues that businesses should move beyond relying on prompts alone. Yoav says the term distracts from understanding AI’s specific strengths and weaknesses. He identifies high costs and unreliable outputs as barriers to enterprise deployment, arguing that systems combining LLMs with retrieval, tools and traditional code can manage cost and improve reliability. His proposed approach for startups is “product-algo fit”: design products to use AI systems’ strengths while accounting for their imperfections. He stresses that uncertainty in LLM calls, search and retrieval will not completely go away. The excerpt does not provide measurements showing how much these systems improve cost or reliability, or establish how well the approach works in deployed products.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
Labs · Article
Source details
Original source date: Feb 27, 2025 (2025-02-27T12:09:32Z).
Date basis: Date from the publisher’s own records, through its API.
Added to Escape Meridian: 2026-11-15T02:58:23.301Z. First observed: 2026-10-25T06:57:51.738Z. Retrieved: 2026-10-25T06:59:47.654Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 25, 2026; originally published February 27, 2025. The excerpt identifies its contributor only as Yoav; it does not provide a surname. The excerpt offers no measurements or comparison establishing the scale of cost or reliability improvements from AI systems, or evidence of their performance in deployed products.
Open question
What evidence from deployed products would show whether combining with retrieval, tools and traditional code improves reliability and cost?
Source-stated
Yoav argues that treating AGI as a discrete point distracts from discussing AI technology’s specific strengths and weaknesses.
I believe this loose thinking and hype around AGI is a distraction. Instead, let’s talk concretely about AI technology, its strengths and weaknesses.
Source-stated
Yoav identifies high LLM operating costs and unreliable outputs as challenges that keep many enterprise AI projects from deployment.
Cost: The high cost of running LLMs challenges the economics of business software.
Source-stated
Yoav argues that AI systems combining LLMs with retrieval, tools and traditional code can offer greater control, efficiency and reliability.
The industry will realize that LLMs are an element of a more comprehensive AI system, which can seamlessly integrate and leverage the strengths of various AI technologies, including LLMs, retrieval, tools and other traditional code.
Source-stated
Yoav advises startup leaders to pursue “product-algo fit” by designing products around AI systems’ strengths and imperfections.
understand the strengths and weaknesses of the technology, and craft that product in a way that leverages their strengths and compensates for their imperfections.
Source-stated
Yoav says uncertainty in LLM calls, search and retrieval will not completely go away.
the underlying uncertainty involved in LLM calls, search and retrieval will not completely go away.
Source and contributor context
AI21 Labs · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Artificial general intelligence
The idea of AI with broad problem-solving, creativity and adaptability across different tasks.
Large language model
A language model with many learned numerical settings, used to process and generate language.
AI21
AI21 publishes technical research and product announcements about its language models and AI systems.
Editorial provenance
Version 1. Publication ID: 1ae3ad7d-f46a-bd20-6945-bcd673383e7e.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-15T02:54:12.859Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: 489bc277c4b2e901e8b4bd6f93f797c87accb9f34a67ad02aa44278fdb729ee0. Source text hash: f584e48e6b3b9eb600b30afb475d882faa4e19e65a6d1110c9b83169d78e5a54.
Revision and source-check history
Version 1 · 2026-11-15T02:58:23.301Z
Yoav argues in AI21 Labs’ excerpt for building complete AI systems
Published February 27, 2025, this excerpt from Yoav’s contribution to Google Cloud’s “Future of AI: Perspectives for Startups 2025” report argues that businesses should move beyond relying on prompts alone. Yoav says the term distracts from understanding AI’s specific strengths and weaknesses. He identifies high costs and unreliable outputs as barriers to enterprise deployment, arguing that systems combining LLMs with retrieval, tools and traditional code can manage cost and improve reliability. His proposed approach for startups is “product-algo fit”: design products to use AI systems’ strengths while accounting for their imperfections. He stresses that uncertainty in LLM calls, search and retrieval will not completely go away. The excerpt does not provide measurements showing how much these systems improve cost or reliability, or establish how well the approach works in deployed products.
Original source
Labs · Article
Source details
Original source date: Feb 27, 2025 (2025-02-27T12:09:32Z).
Date basis: Date from the publisher’s own records, through its API.
Added to Escape Meridian: 2026-11-15T02:58:23.301Z. First observed: 2026-10-25T06:57:51.738Z. Retrieved: 2026-10-25T06:59:47.654Z.
Limits
Text as retrieved on October 25, 2026; originally published February 27, 2025. The excerpt identifies its contributor only as Yoav; it does not provide a surname. The excerpt offers no measurements or comparison establishing the scale of cost or reliability improvements from AI systems, or evidence of their performance in deployed products.
Open question
What evidence from deployed products would show whether combining with retrieval, tools and traditional code improves reliability and cost?
Source-stated
Yoav argues that treating AGI as a discrete point distracts from discussing AI technology’s specific strengths and weaknesses.
I believe this loose thinking and hype around AGI is a distraction. Instead, let’s talk concretely about AI technology, its strengths and weaknesses.
Source-stated
Yoav identifies high LLM operating costs and unreliable outputs as challenges that keep many enterprise AI projects from deployment.
Cost: The high cost of running LLMs challenges the economics of business software.
Source-stated
Yoav argues that AI systems combining LLMs with retrieval, tools and traditional code can offer greater control, efficiency and reliability.
The industry will realize that LLMs are an element of a more comprehensive AI system, which can seamlessly integrate and leverage the strengths of various AI technologies, including LLMs, retrieval, tools and other traditional code.
Source-stated
Yoav advises startup leaders to pursue “product-algo fit” by designing products around AI systems’ strengths and imperfections.
understand the strengths and weaknesses of the technology, and craft that product in a way that leverages their strengths and compensates for their imperfections.
Source-stated
Yoav says uncertainty in LLM calls, search and retrieval will not completely go away.
the underlying uncertainty involved in LLM calls, search and retrieval will not completely go away.
Source and contributor context
AI21 Labs · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Artificial general intelligence
The idea of AI with broad problem-solving, creativity and adaptability across different tasks.
Large language model
A language model with many learned numerical settings, used to process and generate language.
AI21
AI21 publishes technical research and product announcements about its language models and AI systems.
Editorial provenance
Version 1. Publication ID: 1ae3ad7d-f46a-bd20-6945-bcd673383e7e.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-15T02:54:12.859Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: 489bc277c4b2e901e8b4bd6f93f797c87accb9f34a67ad02aa44278fdb729ee0. Source text hash: f584e48e6b3b9eb600b30afb475d882faa4e19e65a6d1110c9b83169d78e5a54.
passed · 2026-11-15T02:54:12.859Z · anthropic/claude-opus-5.5
Publication: 1ae3ad7d-f46a-bd20-6945-bcd673383e7e
attribution — pass: The reading attributes the argument to Yoav, matching the excerpt's identification, and places it on AI21 Labs' page as an excerpt from a Google Cloud report. It correctly notes in limitations that no surname is given. Claims stay framed as Yoav's assertions.
Below is an excerpt from Yoav’s contribution to Google Cloud’s ‘Future of AI: Perspectives for Startups 2025’ report.
summary — pass: The summary captures the central thesis: AGI talk is a distraction, cost and unreliability block enterprise deployment, and composite AI systems plus product-algo fit are the answer. LLM and AGI are defined on the page, and product-algo fit is explained. Some of the author's force is softened, including the prompt and pray phrase, the 95%/5% example and his view that hoping hallucination disappears is delusional wishful thinking. These omissions do not distort the position. The question is substantive and grounded in the source.
it’s delusional wishful thinking that this will go away, no matter how much effort is placed on things like “alignment”, “guardrails”, and such
date — pass: February 27, 2025 matches the UTC day of originalPublished (2025-02-27T12:09:32Z) at appropriate precision. The kind is publisher-page, so no archive disclosure is needed. The reading uses no later events or hindsight. Forecasts such as AI systems offering better reliability are framed as Yoav's arguments. The page was modified in 2026, but the cited passages are plainly part of the coherent original excerpt.
support — pass: Each claim's excerpt supports it. Yoav links deployment barriers to cost and nonsense output, and the source also states that only a fraction of projects reach deployment. The AI systems benefit is presented as his argument and not as a demonstrated result, and the limitations correctly note the absence of measurements. The page was edited after publication (dateModified 2026-03-25), but all cited text is integral to the original argument, so no claim appears to rely on a later addition.
But for all of the mass experimentation going on in enterprise, only a fraction of AI projects actually reach deployment. This boils down to two key challenges:
media — pass: No media was selected and no verified candidate media were supplied, so there is nothing to misidentify.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
AI21 publishes technical research and product announcements about its language models and AI systems.
Explore AI21The idea of AI with broad problem-solving, creativity and adaptability across different tasks.
Definition referenceA language model with many learned numerical settings, used to process and generate language.
Definition referenceAI21 publishes technical research and product announcements about its language models and AI systems.
Explore AI21A language model with many learned numerical settings, used to process and generate language.
Definition referenceThe idea of AI with broad problem-solving, creativity and adaptability across different tasks.
Definition referenceA language model with many learned numerical settings, used to process and generate language.
Definition referenceAI21 publishes technical research and product announcements about its language models and AI systems.
Explore AI21A language model with many learned numerical settings, used to process and generate language.
Definition reference