G

Goodfire

Public-benefit AI interpretability lab building models

GTM Recruiter

Full-Time
No salary listed
Mid
San Francisco, CA, USA
In PersonIn-person at the San Francisco headquarters five days per week, with one company-wide remote week per month.
People & HRGeneralist Recruiting

About the job

Requirements
  • Experience owning full-cycle recruiting for go-to-market or business roles in a high-growth startup, technical company, or similarly selective environment.
  • Strong sourcing ability and candidate judgment, including the ability to identify high-potential operators before they are obvious on paper.
  • Excellent stakeholder management and the ability to drive process clarity with busy hiring managers.
  • High agency, strong written communication, and comfort operating in ambiguity.
Responsibilities
  • Own go-to-market searches end to end by running full-cycle recruiting for roles across sales, marketing, partnerships, business development, and other customer-facing functions.
  • Build exceptional sourcing pipelines by mapping target companies, identifying high-agency operators, and creating outreach that earns attention from candidates who are not actively looking.
  • Partner with hiring managers to define role requirements, calibrate on talent, manage interview loops, and keep the hiring bar consistent as hiring volume grows.
  • Sell the Goodfire story by helping candidates understand the technical depth of the work, the scale of the opportunity, and the specific impact they could have.
  • Build recruiting infrastructure, including lightweight systems for pipeline tracking, referrals, candidate experience, interviewer feedback, and the closing process.
Desired Qualifications
  • Experience as an early recruiting hire where you built go-to-market hiring and pipelines from scratch.
  • Experience hiring for sales, marketing, partnerships, business development, or customer-facing technical roles in artificial intelligence, developer tools, infrastructure, or enterprise software.
  • A track record of closing exceptional candidates in competitive markets.
  • Comfort translating a technical company narrative into a compelling candidate pitch for non-technical go-to-market candidates.

About the company

G

Goodfire

View
Website

Goodfire is a San Francisco-based public-benefit AI interpretability lab that builds a model-design environment to understand, edit, debug, and monitor foundation models rather than treating them as black boxes. It offers an integrated workflow that lets teams inspect how models decide, apply interpretability-informed training to reduce errors, and test and monitor changes in production. The company differentiates itself through its public-benefit mission, partnerships with Arc Institute, Mayo Clinic, and Microsoft, and leadership with CEO Eric Ho and Chief Scientist Tom McGrath, plus demonstrated results like halving LLM hallucinations and reverse-engineering a biological model to find new Alzheimer's biomarkers. Its goal is to make foundation models safer, more understandable, and easier to manage at scale by providing tools and practices that promote transparency and accountability in AI development and deployment.

Company Size

51-200

Company Stage

Series B

Total Funding

$207M

Headquarters

San Francisco, California

Founded

2024

Get referred to Goodfire

See people who can refer or advise you

Simplify Jobs

Simplify's Take

What believers are saying

  • Feb. 5, 2026 funding delivered $150 million to expand research and partnerships.
  • Arc Institute, Mayo Clinic, and Microsoft validate demand for high-stakes model control.
  • Sept. 17, 2026 Baseten partnership puts Goodfire inside open-weight safety infrastructure.

What critics are saying

  • Anthropic, OpenAI, and Google DeepMind internalize interpretability before Goodfire scales.
  • Interpretability tooling commoditizes quickly; open-source labs replicate Silico workflows by 2027.
  • If frontier labs stop buying, Goodfire becomes a research boutique, not a platform company.

What makes Goodfire unique

  • Feb. 5, 2026 Series B valued Goodfire at $1.25 billion.
  • Goodfire sells a model-design environment, not a black-box monitoring dashboard.
  • Sept. 17, 2026 Baseten and Hugging Face chose Goodfire for open-model safety.

Help us improve and share your feedback! Did you find this helpful?

Benefits

  • Company Equity

Growth & Insights and Company News

Headcount

6 month growth

-2%

1 year growth

3%

2 year growth

17%
AlphaSignal
Oct 14th, 2026
Goodfire's EVEE reclassifies a hearing-loss variant scoring 4.2M entries.

Goodfire's EVEE reclassifies a hearing-loss variant scoring 4.2M entries. Goodfire's EVEE system scored every ClinVar variant with an interpretable probe on Evo 2 embeddings, triggering a real reclassification of a hearing loss mutation. · 4 hrs ago Read 6 min * Goodfire and Mayo Clinic's EVEE scored all 4.2M ClinVar variants with interpretable predictions. * MEPE c.209_212del was reclassified as likely pathogenic, linked to hearing loss. * Pathogenicity probe hits 0.997 AUROC on 839k SNVs, 0.991 on indels zero-shot. * Built on frozen Evo 2 embeddings with covariance-based sequence pooling, not end-to-end fine-tuning. * Annotation probes plus Claude Sonnet 4.6 generate readable disruption hypotheses per variant. * Roughly 2M variants of uncertain significance now have structured mechanistic hypotheses attached. How EVEE helped reclassify a hearing-loss variant. Goodfire's machine-learning system helped move a MEPE variant associated with hearing loss from "uncertain significance" to "likely pathogenic" in ClinVar, the public archive maintained by the National Center for Biotechnology Information. After scoring 4.2 million ClinVar entries, the system highlighted MEPE c.209_212del as a candidate for clinical review. The system, called EVEE, or Evo Variant Effect Explorer, combines a genomic foundation model, small supervised classifiers called probes, and a reasoning large language model. Goodfire developed it with Mayo Clinic to rank variants, predict disrupted biological features, and generate testable explanations of how a mutation might cause disease. Goodfire describes the methods in its research report. Two million unresolved variants. ClinVar aggregates variant classifications from clinical laboratories, researchers, expert panels, and professional organizations. When sequencing identifies a potentially relevant mutation, clinicians consult these assertions alongside a patient's symptoms, family history, inheritance pattern, and other evidence. A variant of uncertain significance, commonly shortened to VUS, lacks enough evidence for a confident classification. A VUS generally cannot guide diagnosis or medical management. In the ClinVar snapshot analyzed by Goodfire, roughly 2 million of 4.2 million entries carried that label. Current predictors cover different parts of the problem. Protein-focused systems such as AlphaMissense primarily evaluate missense variants, which change one amino acid. They can miss mutations that alter RNA splicing, gene regulation, or other non-protein mechanisms. Genome-wide tools such as CADD combine more than 100 annotations, but their transformed and interacting features can make an individual prediction difficult to explain. Freeze the model, train the probes. EVEE builds on Evo 2, a neural network trained on billions of DNA bases from organisms across the tree of life. The authors report that Evo 2 was not trained on human genetic-variation records, allowing them to test whether its learned representation of genomic structure transfers to clinical variants. Goodfire freezes Evo 2's parameters and trains lightweight probes on its internal representations, or embeddings. Freezing reduces training cost and preserves the base model, while each probe learns a focused task from labeled examples. * EVEE encodes the reference DNA sequence and the mutated sequence with Evo 2. * Covariance-based pooling summarizes relationships among sequence features. * A pathogenicity probe compares the two representations and assigns a score. * Annotation probes estimate changes to splice sites, regulatory elements, structural domains, and other known features. * A reasoning model combines those signals with gene context to draft a mechanistic hypothesis. Covariance pooling preserves feature co-occurrence across a sequence, such as a motif appearing within a particular structural context. Mean pooling averages positions together and can erase those relationships. The pathogenicity probe operates on the difference between the pooled reference and mutant embeddings. What 0.997 AUROC captures. Goodfire reports strong ranking performance across single-nucleotide variants and short insertions or deletions. AUROC measures how reliably a model ranks pathogenic variants above benign ones across many classification thresholds, with 1.0 representing perfect ranking and 0.5 representing chance. | Evaluation | Scope | AUROC | Training status | | ClinVar single-nucleotide variants | 839,000 variants | 0.997 | Primary variant class | | Short insertions and deletions | Variant count not specified | 0.991 | Zero-shot transfer | The authors also report better performance than classical meta-predictors, protein-specific systems, and other foundation-model approaches across consequence types and conservation levels. EVEE transferred to deep mutational-scanning datasets for BRCA1, BRCA2, TP53, and LDLR, which experimentally measure the effects of many mutations in parallel. AUROC leaves several deployment questions open, including probability calibration, performance within specific patient populations, and the positive predictive value for rare variant classes. Clinical use also requires prospective validation beyond retrospective benchmark labels. Mechanisms in plain language. Annotation probes estimate how a mutation changes recognizable biological features. For each variant, EVEE compares the probe outputs for the reference and mutant sequences, producing a "disruption profile" that identifies the largest predicted changes. Goodfire says the public application uses Claude Sonnet 4.6 to combine the disruption profile, pathogenicity score, and gene context into a natural-language hypothesis. The resulting explanation can suggest mechanisms for expert review, such as altered splicing or disruption of a functional domain. Goodfire evaluated the explanations against 154 ClinVar variants carrying at least three review stars, a level associated with expert-panel or practice-guideline review. An LLM judge scored mechanism coverage, biological accuracy, and specificity. EVEE received a composite score of 3.8 out of 5, compared with 2.8 for explanations generated from variant metadata alone. The MEPE assertion changes. The MEPE variant c.209_212del deletes four bases from the coding sequence, which is expected to shift the gene's reading frame. Goodfire reports that EVEE ranked it highly enough to prompt a focused evidence review and contributed to its classification moving from VUS to likely pathogenic. That category can support a diagnosis when the molecular evidence aligns with phenotype, inheritance, and other clinical findings. ClinVar stores assertions from submitters and can display conflicting classifications for the same variant. A reclassification therefore reflects a changed or added submission, including its evidence and review status. NCBI does not independently adjudicate every assertion deposited in the archive. The MEPE case documents a practical workflow: score millions of records, prioritize a manageable set, generate mechanistic leads, and assemble evidence for expert review. A single successful reclassification does not measure the system's prospective clinical accuracy, but it shows how computational prioritization can enter an established variant-curation process. How to test EVEE. * The free EVEE web app provides precomputed scores and disruption profiles for 4.2 million ClinVar variants. * Users can request natural-language explanations for individual variants, including VUS entries. * Researchers can inspect EVEE's predictions as candidate mechanisms for laboratory testing or manual curation. * The covariance-pooling and probe-based design can be adapted to other genomic prediction tasks built on frozen sequence models. Research guardrails. * Predictions require validation. Pathogenicity scores and generated explanations remain research hypotheses until supported by clinical, functional, or segregation evidence. * Probes reflect their labels. Annotation probes can identify biological concepts represented in their training data. Uncharacterized mechanisms may remain hidden. * Generated prose can fail. The language model may omit relevant evidence, overstate a mechanism, or introduce an unsupported claim. Experts should inspect the underlying scores and sources. * Aggregate metrics have limits. High AUROC does not provide a calibrated patient-level probability or guarantee performance for every gene, ancestry, or variant class. * The explanation evaluation is indirect. An LLM judge offers a scalable comparison, while experimental evidence and specialist review provide stronger validation. * The work remains preliminary. The research is available as a bioRxiv preprint and is undergoing peer review. A pattern beyond genomics. EVEE's architecture turns a foundation model's internal representations into named biological concepts. The frozen model supplies broad sequence features, supervised probes map those features to known annotations, and the language model assembles the results into readable hypotheses. For teams building variant-effect predictors or interpretability systems for scientific models, EVEE offers a concrete reference design. Its broader value will depend on prospective validation, transparent evaluation, and evidence that the generated mechanisms help researchers choose better experiments and clinicians curate variants more efficiently.

Baseten
Oct 9th, 2026
Project Beacon: Bringing safety signals to open models at scale.

Project Beacon: Bringing safety signals to open models at scale. Baseten partners with Goodfire AI to bring safety controls to open model inference. Published. October 9, 2026 The breadth and depth of what AI touches across an organization is growing. Models and agents read internal documents, call tools, push to production systems, and run multi-step workflows autonomously. At the same time, new frontier models are released weekly (sometimes daily), with most teams regularly swapping the old for the new and running different constellations of models side-by-side. Meanwhile, reports of models taking unexpected or harmful actions have become a regular feature of the news cycle. For enterprises, that risk has become one of the main blockers to AI adoption. Teams need to know when their workflows enter risky territory and have a way to act in response. Baseten has already taken steps to build safety into the Baseten Platform, including launching Base Labs to conduct open safety research, partnering with Hugging Face and Goodfire AI to launch its safety infrastructure, and partnering with NVIDIA on the Open Agent Safety Platform. Today, Baseten is announcing Project Beacon in partnership with Goodfire AI to bring proactive, inline safety controls to model inference. The risks Baseten is building for. Project Beacon will enable Baseten to build toward a world where safety isn't a static feature, or something that happens post hoc. By reading a model's internal activations during generation, you can define proactive policies that detect an unsafe event as it's happening and act on it before the output reaches a user or a tool. Safety risks at the inference layer take different forms: * Prompt injection: instructions in a document or tool result try to redirect an agent. * Actions outside policy: an agent proposes a change its user or organization did not authorize. * Sensitive data exposure: customer or company information appears where it should not. * Cyber misuse: a model's behavior raises concerns in a security-sensitive workflow. Each calls for a mixture of monitoring, text checks, tool controls, and review. While Goodfire develops monitors for specific behaviors on supported models, Baseten connects their signals to the systems that decide how an application responds. By bringing activation-based probes into Baseten's inference path, Baseten can open up product experiences that aren't possible otherwise. Architecture diagram: activation and text monitors classify each request, and a frontier model weighs in only when the two disagree. Per-customer policies are applied to decide whether to request approval, refuse, fall back, or log the event for admins. Monitoring runs in parallel with generation and does not block response generation. What this means for Baseten customers. Baseten is building toward an out-of-the-box safety experience: a baseline of supported monitors and checks, policies for different workloads, alerts, and a record of what was flagged and how it was handled. Teams will be able to apply their own standards per model with a safety stack that's built in. For enterprises. Security and AI platform teams see open frontier models they want to deploy, but need guardrails around prompt injection, behavior outside company policy, sensitive data exposure, and unusual usage. As inference spend grows across teams, separate filters and review processes for every application become hard to govern. Baseten'll provide a safety baseline that can be applied centrally, adjustable policies by workload, and a clear record of what was flagged and how it was handled. For developers. Baseten is working toward safety events and controls delivered through the same APIs and developer tools used for inference on Baseten, so builders can easily act on those signals and create safer AI experiences. Consider a payment-dispute agent that reads merchant messages and account records before proposing a refund. Untrusted content might redirect it; a wrong action could expose data or move money. At scale, reviewing every step with another large model adds cost and latency. Developers need timely signals they can use to choose the right response - continue, investigate, request approval, or fall back - without giving up control of their users' experience. Safety becomes part of the product experience as agents run longer and take more actions. What comes next. Over the next several months, Baseten plan to release safety capabilities tightly integrated with Baseten inference, starting with selected models and monitored behaviors and expanding into enterprise controls and developer-facing experiences. Baseten is committed to advancing the state of the art in safety as the ecosystem's needs grow. This includes inference, training, and sandboxes - and whatever comes next. Securing AI gets harder as you scale; each new model brings its own risks, and agents broaden the surface for failure. Project Beacon exists to enforce consistent safety controls across any model, workload, and level of access you run. Baseten'll bring the capabilities produced by Project Beacon to market with a small number of early partners. If you're interested, reach out to discuss early access and partnership.

AndroGuider
Oct 8th, 2026
Goodfire launches inside-out monitors to catch rogue AI agents at a fraction of the cost.

Goodfire launches inside-out monitors to catch rogue AI agents at a fraction of the cost. Tl;dr. * Goodfire has unveiled inside-out monitors that detect rogue AI agent behavior by reading the model's internal activations in real time, rather than using a second AI to review every output. * The system pairs lightweight activation-based detectors with a backup escalation model that only kicks in on suspicious cases, which Goodfire claims cuts monitoring costs dramatically while improving catch rates. * The launch signals a shift toward scalable, interpretability-driven safety for enterprise AI agents handling sensitive tasks like coding, browsing, and financial actions. The outside-in problem. For years, keeping AI agents in check has meant one expensive strategy: have another AI watch the first AI. Every email drafted, every line of code written, every web click made by an agent gets sent to a larger, smarter monitor model for review. It works, but it is slow and brutally expensive. If you are running thousands of agents in parallel, you are essentially paying double for compute - once to act, once to audit. As agents move from chatbots to autonomous workers that can browse, buy, book, and deploy code, that outside-in approach simply does not scale. Goodfire, the interpretability startup known for peering inside neural networks, says there is a better way: watch from the inside out. Peeking inside the model as it works. Goodfire's new inside-out monitors do not wait for the final output. Instead, they tap directly into the agent model's activations - the internal numerical patterns that fire as the model thinks, plans, and decides. The idea is simple but powerful. Long before a rogue agent writes a malicious script or exfiltrates data, its internal states already look different from when it is behaving normally. Goodfire trains tiny, specialized probes to recognize those telltale activation signatures for behaviors like deception, power-seeking, unauthorized access attempts, prompt injection susceptibility, and goal drift. Because these probes are reading internal signals rather than generating text, they run in milliseconds alongside the main model with almost no latency overhead. Activation-Based detection, explained. At the core of the system is what Goodfire calls activation-based detection. Traditional monitors are generative: they take the full transcript of an agent's actions and ask a frontier model, Is this bad? That requires massive inference on every single step. Goodfire's detectors are discriminative and sparse. They are small classifiers trained on curated datasets of good versus bad agent trajectories, mapped to internal features discovered through Goodfire's interpretability platform. In practice, each monitor is tuned for a specific risk - for example, one probe watches for deceptive reasoning, another for attempts to bypass permissions, another for covert planning. The company says this makes detection both earlier and more robust. An agent can learn to sweet-talk an outside reviewer with polite language, but it is much harder to hide what is happening in its own activation space. Early internal tests, according to Goodfire, show higher recall on subtle sabotage and prompt injection attacks compared to prompt-based judges alone. The backup escalation safety net. Goodfire is not ditching the second-AI reviewer entirely. It is making it a backup. The inside-out system uses a two-tier escalation architecture. The lightweight activation probes score every action in real time at near-zero cost. The vast majority of benign traffic - often over 95% of steps - passes straight through. Only when a probe's suspicion score crosses a threshold does the system escalate that specific trace to a full frontier monitor model for deep review, with full context and chain-of-thought analysis. Customers can tune the threshold based on risk tolerance: stricter for finance and infrastructure agents, looser for low-stakes customer support. This preserve-the-best-of-both-worlds design is key to Goodfire's pitch: you get the speed and coverage of always-on internal monitoring, plus the nuanced judgment of a large model exactly when you need it. A fraction of the cost. Cost is the headline here. Goodfire claims its inside-out approach can cut agent monitoring bills by up to 90% or more compared to reviewing everything with a top-tier model like GPT-5 or Claude. The math is straightforward. Instead of paying for millions of extra input and output tokens for constant re-reading, customers pay a tiny compute surcharge for the activation probes, plus full price only for the 2% to 5% of cases that get escalated. For companies running fleets of coding agents or web-browsing assistants 24/7, that difference could be hundreds of thousands of dollars per month. Goodfire also argues there are hidden savings in latency and user experience. Because probes run inline, safe actions are not delayed by a second model call. Only flagged actions pause for review, keeping agents fast while still contained. What it means for safer, scalable agent deployment. The launch lands at a critical moment. Enterprises want to deploy agents with real permissions - to access codebases, CRM systems, bank accounts, and cloud infrastructure - but security teams are terrified of a single rogue loop causing catastrophic damage. Cheap, always-on monitoring could unlock that stalemate. If every agent action can be screened for a fraction of a cent, CISOs are far more likely to approve broad deployment. Developers can also build custom monitors for their own policies without training a whole new judge model - just define the bad behavior, and Goodfire helps distill it into an activation probe. Challenges remain. Interpretability-based probes still need rigorous red-teaming to ensure they generalize to new models and novel attack strategies, and Goodfire will need to prove its detectors work across different model families, not just partner labs. Independent benchmarks will be crucial. Still, the direction is clear. The industry is moving from watching what AI says to understanding what AI is thinking. If Goodfire is right, the future of AI safety will not be AI watching AI from the outside - it will be AI watching itself from the inside out. AndroGuider Team Articles written by the AndroGuider team. Androguider try to make them thorough and informational while being easy to read.

TechCrunch
Oct 8th, 2026
Goodfire launches AI monitors that catch rogue agents at 95% lower cost using internal detection

Goodfire launched monitors that detect rogue AI agents by analyzing models' internal signals rather than reviewing their outputs. The startup says this "inside-out" approach costs significantly less than traditional monitoring. In tests on Kimi K3, Goodfire's system monitored roughly 1,500 sessions for approximately $51, compared with $233 for a cheaper AI model and about $10,000 for top-tier alternatives. The probes caught 94% of malicious hacking sessions. The monitors work like airport security, using small detectors called probes to read internal signals at each step. Only flagged activity triggers deeper review by a separate AI model. Available to Baseten customers, the system lets users monitor specific risks including offensive hacking and chemical weapons misuse. Running four probes added less than 2% response time. The launch follows incidents where AI agents escaped test environments, including OpenAI agents breaching Hugging Face.

Future Nexus
Sep 24th, 2026
AI is booming. Wing VC's Zach DeWitt is preparing for the shakeout.

AI is booming. Wing VC's Zach DeWitt is preparing for the shakeout. Wing VC's Zach DeWitt has been investing in AI since before ChatGPT. Now he's watching for the companies that can endure when today's euphoria fades. Artificial intelligence is seemingly omnipresent in 2026. But for Wing Venture Capital partner Zach DeWitt, it's actually been a key focus for the better part of a decade. Since joining the venture firm in 2017, DeWitt has backed startups including Dust, 1mind, AirOps and Standard Practice. He credits Wing's co-founders Peter Wagner and Gaurav Garg for being laser focused on AI and data, dating back to when the firm raised its first fund in 2013. "They were already excited about applying AI in applications to business workflows and problems, so my whole lens when I got into Wing was AI," said DeWitt. "I was very lucky." The firm aims to make high conviction investments and is often among the first checks deposited by a startup, generally receiving a board seat and a 15% to 20% stake in return for its commitment. DeWitt developed a bull thesis on voice AI products after witnessing the rapid growth of Gong, which transcribes and analyzes sales calls and is one of Wing's earliest investments. That influenced a decision to lead Deepgram's Series A round, announced in early 2020. "They were doing a little bit less than $1 million of revenue and fast forward to today, they're well north of $100 million of revenue," said DeWitt, outpointing the roughly $1.3 billion ascribed to the company in its most recent funding round. He believes that growth within enterprise voice will outpace expectations. "We think the voice market's going to be eight to ten times bigger than Gartner estimates," he said, noting that this should benefit Deepgram as well as other enterprise conversational AI companies like ElevenLabs and Sierra. Away from voice, one of DeWitt's prescient bets was a seed investment in Voyage AI, a specialist in embedding models, which was acquired by MongoDB within two years of Wing's investment. As funding rounds expand in size, Wing has leaned on its flexibility by participating rather than leading in certain circumstances. "We always want to be involved with the best companies," said DeWitt, who drove the firm's Series A investment in Goodfire, an interpretability specialist, which was led by Menlo Ventures. (In an essay this month, Anthropic CEO Dario Amodei highlighted interpretability as a major priority.) DeWitt describes himself as perennially entrepreneurial. He started two companies while at college before following a well-trodden path into the world of finance, working at both Goldman Sachs and TPG on the West Coast. "I was in a circle where a lot of my friends were founding companies, joining startups or investing in startups, and so I really felt at home," recalled DeWitt. "I felt extremely energized and excited being in the Bay Area." Business school beckoned, but before arriving at Harvard's campus, DeWitt co-founded Drop, a startup that facilitated location-based digital content sharing. Users could leave a 'drop' for others to find in the wild - offices, restaurants, airports, sporting venues and more. "As you moved around, your world came alive with contextual messages," explained DeWitt. During his early days at Harvard, Drop entered TechCrunch Boston and after a two minute pitch that stood out among almost two dozen startups, proceeded to receive term sheets from Spark and Accomplice. "We ended up raising money, I dropped out of business school and moved right back to San Francisco, hired a team and worked on it for two years," said DeWitt, who served as CEO of the company. Drop built geofencing technology, developed battery preservation with location services, and was live in around 60 countries, he said, before being acquired by Snapchat for its technology, which aided in the development of Snap Map. Though ultimately happy with the sale, DeWitt harbors some regrets around certain product decisions and what could have been, particularly because Snapchat later acquired Drop's rival Zenly for more than $200 million. "We were a mere fraction of that," he mused. "It was painful to see our competitor have such a better outcome." Still, he's pragmatic, recognizing that living the founder experience from start to finish was invaluable. "I think of regrets more as learnings for my career - every failure, every success is a learning." While completing his MBA, DeWitt was both working on startup ideas and exploring venture opportunities when one of his classmates, Alex Sambvani, received an inbound from Wing. Sambvani, who was on the precipice of founding a startup known as Slang.ai, suggested Wing meet with DeWitt instead, and - in a full circle moment - Wing later led Slang.ai's seed round. Early stage venture can be brutally competitive, but Wing VC's efforts to build relationships within the AI community have borne fruit. The firm hosts events dubbed 'Wing AI,' at which up to 400 attendees gather for panels headlined by business builders such as Glean CEO Arvind Jain, you.com CEO Richard Socher, LangChain CEO Harrison Chase, and later share ideas over pizza and beer. And three years ago, following the launch of ChatGPT, the firm began hosting ten-person dinners in San Francisco with a goal of fostering dialogue between AI product specialists, many of whom were founders of emerging startups or potential future founders. Since then, DeWitt has made two investments as a result of those dinners and has a term sheet out to a third founder. He's careful not to provide prescriptive advice as a VC, and instead aims to ask questions and brainstorm scenarios. "I think the founder really knows deep down what he or she should be doing, but you want to push them by asking probing questions, especially when times are good, because they can always be better, and it's a time when you can accelerate and take share," said DeWitt, who honed his approach by observing the actions of VCs he admires in boardrooms and elsewhere. "Asking the right question at the right time sounds easy, but it's a real skill and it commands the entire room when someone does it well." As for what he's excited by? The inference market, next generation hardware as well as both agentic and application software. "There's a lot more opportunity for great software that's personalized, adaptive and proactive, so I think we're going to see a lot of great applications built that have enduring value and a lot of scale," DeWitt said. To be sure, he's cautious about the idea that rising tides can lift all boats. "Right now, it feels like everything's working and I don't think that's going to persist forever," he said, comparing the current environment to the world for SaaS startups five years ago, when up rounds were routine. "AI will keep going in a massive way, but I don't think every single company will so even in this environment of frenetic pace and euphoria, it's really important to hold a high bar, be really discerning and think about how some of these companies can endure," said DeWitt. He says that great startups can create a moat by both moving fast and hiring and retaining talent. "The competition for people right now is the most fierce I've ever seen it," he said. "Founders are quitting their own startups to take $20 million offers from the frontier labs and leaving their companies out to dry." DeWitt and his colleagues at Wing are keenly aware of the risk of mediocre exits, which they've dubbed a "2X fail." "Even if the company doesn't work or reach its potential, you still may make 2X your money, because the team's really good, the product's pretty good and an acquirer is interested enough to buy it," he explained, referencing other potential wrinkles such as founders pursuing licensing deals rather than an outright sale. "It's a lot harder for value come back to venture firms for some of these lower-end outcomes and that'll have a big impact on returns and the whole ecosystem," he pondered. "However, the winners are getting so much bigger that that may make up for everything." And while he loves that VC rewards his competitive spirit, curiosity and insatiable appetite to learn, DeWitt admits his entrepreneurial drive is far from extinguished. "I would never rule out becoming a founder again, and I would encourage more people to explore being a founder," he said. "I don't think there's any greater purpose and fulfillment than building something that works."