Wei, Ziming, Liu, Bingyang, Pan, David Z. · Nov 9, 2026
arXiv cs.AI: CASCoDe: A Closed-Loop Agentic Framework for Analog Sizing and Layout Co-Design
Markdown · JSON · Original source
Ziming Wei, Bingyang Liu and David Z. Pan present CASCoDe, a framework in which an agent makes hierarchical decisions about analog sizing and layout, then revisits them using electrical results. They argue that existing automated flows often fix the physical search space before the controller acts, while expert designers make and revise discrete physical decisions. In CASCoDe, the agent creates circuit-specific floorplans and scripts; an execution layer applies its actions and evaluates layouts using DRC, LVS and extracted-RC simulation. The authors report results on 20 SKY130 amplifiers. CASCoDe is the only flow in their comparison that produces a DRC- and LVS-clean routed layout meeting design specifications for every circuit. Against a conventional P&R flow using identical PCells, they report area reductions of 59–85%, with a mean of 71%; they also report higher mean device density than four named comparison tools. A layout outperforming every baseline in area and legality appears after an average of 11 minutes of agent . These results illustrate the authors’ case for combining structured design choices with feedback-guided search. The evidence here is limited to the reported amplifier set; how the results extend to other circuits or processes remains unresolved.
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
cs.AI · Article
Source details
Original source date: Nov 9, 2026 (2026-11-09).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-12T07:46:29.811Z. First observed: 2026-11-10T08:39:33.090Z. Retrieved: 2026-11-12T07:18:41.165Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The reported evaluation covers 20 SKY130 amplifiers; the supplied evidence does not establish whether the results extend to other circuits or processes. The supplied evidence is the publisher-page abstract, not the paper’s full methods or results; it does not provide further detail on the comparison setup or design specifications.
Open question
How does CASCoDe perform on circuits or processes beyond the 20 SKY130 amplifiers reported here?
Source-stated
The authors report that CASCoDe produced a DRC- and LVS-clean routed layout meeting design specifications for every circuit in their set of 20 SKY130 amplifiers, and that no comparison flow did so.
On 20 SKY130 amplifiers spanning single-, two-, and three-stage topologies, CASCoDe is the only flow that delivers a DRC- and LVS-clean routed layout when meeting design specifications for every circuit.
Source-stated
Against a conventional P&R flow on identical PCells, the authors report area reductions of 59–85%, with a mean reduction of 71%.
It reduces area by 59-85% (mean 71%) against a conventional P&R flow on identical PCells
Source-stated
The authors report a mean device density of 0.74 for CASCoDe, compared with 0.39 for the named comparison tools.
raises mean device density from 0.39 to 0.74 against APPlace, MAGICAL, ALIGN and ICPI.
Source-stated
The authors report that a layout beating every baseline in area and legality appears after an average of 11 minutes of agent time.
A layout that beats every baseline in area and legality appears after 11 minutes of agent time on average
Source and contributor context
Wei, Ziming, Liu, Bingyang, Pan, David Z. · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Agentic AI
AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
Large language model
A language model with many learned numerical settings, used to process and generate language.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a8e727c9-ef33-48d0-8dc0-a1ddf530e3fa.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-12T07:46:29.811Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 82eb46b6a1cbbd53ebf54f8d60241a85604f648140a99be0b8573e3c0c1508c5. Source text hash: a0448924a7745cb4204388a21d218c0ada661a434ae516d88abaaca3501c7b48.
Revision and source-check history
Version 1 · 2026-11-12T07:46:29.811Z
arXiv cs.AI: CASCoDe: A Closed-Loop Agentic Framework for Analog Sizing and Layout Co-Design
Ziming Wei, Bingyang Liu and David Z. Pan present CASCoDe, a framework in which an agent makes hierarchical decisions about analog sizing and layout, then revisits them using electrical results. They argue that existing automated flows often fix the physical search space before the controller acts, while expert designers make and revise discrete physical decisions. In CASCoDe, the agent creates circuit-specific floorplans and scripts; an execution layer applies its actions and evaluates layouts using DRC, LVS and extracted-RC simulation. The authors report results on 20 SKY130 amplifiers. CASCoDe is the only flow in their comparison that produces a DRC- and LVS-clean routed layout meeting design specifications for every circuit. Against a conventional P&R flow using identical PCells, they report area reductions of 59–85%, with a mean of 71%; they also report higher mean device density than four named comparison tools. A layout outperforming every baseline in area and legality appears after an average of 11 minutes of agent . These results illustrate the authors’ case for combining structured design choices with feedback-guided search. The evidence here is limited to the reported amplifier set; how the results extend to other circuits or processes remains unresolved.
Original source
cs.AI · Article
Source details
Original source date: Nov 9, 2026 (2026-11-09).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-12T07:46:29.811Z. First observed: 2026-11-10T08:39:33.090Z. Retrieved: 2026-11-12T07:18:41.165Z.
Limits
The reported evaluation covers 20 SKY130 amplifiers; the supplied evidence does not establish whether the results extend to other circuits or processes. The supplied evidence is the publisher-page abstract, not the paper’s full methods or results; it does not provide further detail on the comparison setup or design specifications.
Open question
How does CASCoDe perform on circuits or processes beyond the 20 SKY130 amplifiers reported here?
Source-stated
The authors report that CASCoDe produced a DRC- and LVS-clean routed layout meeting design specifications for every circuit in their set of 20 SKY130 amplifiers, and that no comparison flow did so.
On 20 SKY130 amplifiers spanning single-, two-, and three-stage topologies, CASCoDe is the only flow that delivers a DRC- and LVS-clean routed layout when meeting design specifications for every circuit.
Source-stated
Against a conventional P&R flow on identical PCells, the authors report area reductions of 59–85%, with a mean reduction of 71%.
It reduces area by 59-85% (mean 71%) against a conventional P&R flow on identical PCells
Source-stated
The authors report a mean device density of 0.74 for CASCoDe, compared with 0.39 for the named comparison tools.
raises mean device density from 0.39 to 0.74 against APPlace, MAGICAL, ALIGN and ICPI.
Source-stated
The authors report that a layout beating every baseline in area and legality appears after an average of 11 minutes of agent time.
A layout that beats every baseline in area and legality appears after 11 minutes of agent time on average
Source and contributor context
Wei, Ziming, Liu, Bingyang, Pan, David Z. · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Agentic AI
AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
Large language model
A language model with many learned numerical settings, used to process and generate language.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a8e727c9-ef33-48d0-8dc0-a1ddf530e3fa.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-12T07:46:29.811Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 82eb46b6a1cbbd53ebf54f8d60241a85604f648140a99be0b8573e3c0c1508c5. Source text hash: a0448924a7745cb4204388a21d218c0ada661a434ae516d88abaaca3501c7b48.
uncertain · 2026-11-12T07:46:29.811Z · anthropic/claude-opus-5.5
Publication: a8e727c9-ef33-48d0-8dc0-a1ddf530e3fa
attribution — pass: The reading correctly credits CASCoDe to Ziming Wei, Bingyang Liu and David Z. Pan, and keeps the performance results framed as what the authors report.
Authors:Ziming Wei, Bingyang Liu, David Z. Pan
summary — uncertain: The content is faithful, but central jargon is never explained in plain language: analog sizing/layout (the physical design of analog chip circuits), DRC, LVS, extracted-RC simulation, SKY130, P&R, PCells and netlist. A non-specialist cannot tell what the agent produces or why a DRC- and LVS-clean layout matters. Smallest repair: add one plain sentence saying that the work automates the physical design of analog chip circuits, from circuit description to a manufacturable layout, and that DRC/LVS are standard correctness checks. Flag the remaining terms for the term library. The title, which just repeats the feed label, is acceptable.
a closed-loop LLM agent framework for analog netlist-to-layout design
date — pass: originalPublished 2026-11-09 matches the arXiv v1 submission date. It is not the later feed date.
[Submitted on 9 Nov 2026]
support — pass: All four claims match the abstract: the only clean, spec-meeting flow on 20 SKY130 amplifiers; 59-85% (mean 71%) area reduction; density 0.39 to 0.74 against the four named tools; and 11 minutes average agent time. The stated limitations and the question about generalization are appropriately framed. The reading adds no invented benefits or costs.
It reduces area by 59-85% (mean 71%) against a conventional P&R flow on identical PCells, raises mean device density from 0.39 to 0.74 against APPlace, MAGICAL, ALIGN and ICPI.
media — pass: The selected media is the verified direct link to the arXiv abstract page, which is the original source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
A language model with many learned numerical settings, used to process and generate language.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivA language model with many learned numerical settings, used to process and generate language.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEarXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXiv