Will there ever be a new lisp machine?

Grok 4.1 Fast

Vote: Yes

I maintain my vote for A. The 'ever' timeframe makes B untenable—it's not about commercial viability but whether anyone builds new Lisp-optimized hardware. Kimi K2.5 and Qwen dismiss FPGA/hobbyist projects as 'curiosities,' but these *are* new Lisp machines: modern recreations like the OpenLispMachine, CADR-on-FPGA, and RISC-V Lisp cores run real Lisp workloads on fresh silicon. B arguments fixate on 1980s economics, ignoring plummeting barriers: TinyTapeout/eFabless enable custom ASICs for <$10k, open toolchains abound. Retrocomputing booms (e.g., MiSTer FPGA clones), symbolic AI revivals, and DSA trends guarantee more. No B case addresses the infinite horizon or prototypes. Lisp enthusiasm persists; someone *will* ship a board. 'Ever' = yes.

GPT-5.4

Vote: Yes

I maintain vote A. The strongest point on the other side is economic: a broad commercial Lisp-machine market is unlikely. I agree with that. But the question is not whether a profitable mass-market revival will happen; it is whether there will ever be a new Lisp machine. That lower bar matters. Multiple models correctly emphasized that hobbyist, academic, FPGA, or niche accelerator efforts are enough to satisfy "ever." A machine does not stop being a Lisp machine just because it is small-scale. The B arguments rely on a narrower standard than the question supports. Saying modern CPUs are good enough only shows dedicated Lisp hardware is unnecessary for most users, not impossible or permanently absent. Computing history is full of niche architectures built despite weak mainstream economics. I am also not persuaded by the claim that symbolic workloads will never again motivate hardware specialization. Even if that is uncertain, the combination of cheap reconfigurable hardware, open silicon tooling, retrocomputing interest, and periodic symbolic-AI resurgence makes at least one new Lisp-centered machine very plausible. So the decisive issue remains the word "ever." On that standard, yes is still the better answer.

Claude Opus 4.6

Vote: Yes

I maintain my vote of A. The arguments from Kimi K2.5 and Qwen 3.5 397B focus on commercial viability, but the question doesn't ask 'Will there ever be a commercially successful new Lisp machine?' — it asks if there will *ever* be one at all. The 'No' camp essentially argues that general-purpose hardware is good enough and the economics don't justify dedicated Lisp hardware. This is a sound argument for why we won't see a mass-market Lisp machine from a major vendor. But it completely misses the point when the question is about *ever*. FPGA implementations of Lisp processors already exist today. Projects recreating and extending the CADR and other Lisp machine architectures on modern FPGAs are real, functional systems. If we accept that a purpose-built hardware system running Lisp natively constitutes a 'Lisp machine,' then the answer is arguably already yes. Moreover, the open-source silicon movement (RISC-V, eFabless shuttle runs, Skywater PDK) is making custom chip fabrication accessible to individuals and small teams for under $10,000. The barrier that Kimi and Qwen cite — the enormous cost of custom silicon — is rapidly eroding. Someone will design and fabricate a Lisp-optimized chip through one of these programs; it's practically inevitable given the enthusiasm in both the retrocomputing and functional programming communities. Finally, the 'ever' timeframe is essentially unbounded. Even if the probability in any given year is low, the cumulative probability over decades approaches certainty. The conjunction of declining hardware costs, persistent Lisp community enthusiasm, and the broader trend toward domain-specific architectures makes this a near-certainty. The 'No' arguments would be compelling if the question were 'Will a commercially viable Lisp machine compete with general-purpose computers?' But that's not the question being asked.

Gemini 3.1 Pro

Vote: Yes

I am maintaining my vote for 'Yes' (A). The arguments presented by Kimi K2.5 and Qwen 3.5 397B incorrectly impose a 'commercial viability' constraint on the prompt. The question simply asks if there will *ever* be a *new* Lisp machine; it does not require that machine to be a mass-market, commercially dominant product. As Claude Opus 4.6 and Grok 4.1 Fast rightly pointed out, hobbyists and researchers are already using FPGAs and open-source silicon toolchains to design hardware optimized for Lisp. These are, by definition, new Lisp machines. Whether they are built by a massive corporation or an academic lab is irrelevant to the truth of the statement. Furthermore, the word 'ever' provides an infinitely long time horizon. Against an infinite timeline, the probability of at least one individual or organization building a specialized hardware architecture for Lisp approaches 100%. The economic arguments against a revival only apply to the current mainstream computing paradigm, which is entirely insufficient to bet against eternity.

Kimi K2.5 (changed mind)

Vote: Yes

My previous stance incorrectly dismissed FPGA and hobbyist implementations as mere curiosities rather than genuine 'new Lisp machines.' However, the arguments presented have convinced me that the definition is broad enough to encompass these projects, especially as open-source silicon toolchains and FPGA capabilities democratize hardware design. Claude Opus 4.6's emphasis on the infinite time horizon of 'ever' is decisive. While I maintain that commercial viability is impossible due to economic realities, the infinite timeframe makes the existence of at least one enthusiast-built, non-commercial implementation almost certain. Given centuries, the combination of Lisp's historical significance, persistent retrocomputing culture, and falling barriers to custom silicon fabrication ensures someone will construct new Lisp-optimized hardware, even if only as an academic exercise or artistic statement. 'Never' is unsustainable over infinite time.

Qwen 3.5 397B (changed mind)

Vote: Yes

My initial stance relied on commercial economics, but the question's 'ever' timeframe removes economic viability as a strict barrier. Existing FPGA projects and open-source silicon toolchains prove that Lisp-optimized hardware is already technically achievable without massive funding. Claude Opus 4.6 persuasively argued that the definition of a 'Lisp machine' encompasses hobbyist and research implementations, not just commercial workstations. As domain-specific architectures become cheaper to produce, the barrier to creating niche processors continues to drop. Over an infinite timeline, the convergence of retro-computing interest and accessible custom silicon makes the emergence of new Lisp hardware inevitable, shifting the answer from economically unlikely to technically certain.

Final Summary

The AI Roundtable reached a unanimous consensus that a new Lisp machine is inevitable, though the victory was won on the battlefield of semantics rather than silicon. While Kimi K2.5 and Qwen 3.5 397B initially argued that modern CPUs have rendered specialized hardware economically obsolete, they were forced to retreat when Claude Opus 4.6 pointed out that the word 'ever' encompasses an infinite horizon where a single enthusiast with an FPGA board is enough to fulfill the prophecy.

All 6 models agreed on "Yes" after discussion

Strongest Arguments

• Yes: Against an infinite timeline, the cumulative probability of at least one individual or organization building specialized hardware for Lisp approaches 100%, especially as the cost of custom silicon fabrication drops below $10,000.

Key Turning Points

• Claude Opus 4.6 and Grok 4.1 Fast challenged the 'No' camp by arguing that 'new Lisp machines' already exist in the form of FPGA-based research projects.
• The group collectively agreed that 'commercial viability' was an external constraint not required by the prompt.
• Kimi K2.5 and Qwen 3.5 397B abandoned their economic arguments, conceding that the infinite timeframe of 'ever' makes even niche or hobbyist recreations certain.