The frontier is all that matters

I build AI tools, work in product at Chegg, and lead growth at Sapien.

Selected work

• Chegg

  Sulayman Bowles’s product work at Chegg spans Internships.com, student research, recruiting-data pipelines, growth analysis and an AI mock interviewer.

• Sapien

  Sulayman Bowles leads growth and positioning at Sapien, building buyer guides and use cases for AI market research, pricing and consumer insights.

• VOID

  Sulayman Bowles runs VOID, a data brokerage for qualifying data opportunities and permissioned introductions. Earlier consulting work included Atlas.

• Experiments

  Tools and studies in light, texture, and motion.

Selected writing

• Harris County’s toll-road cash

  HCTRA transferred $961 million to other county transportation funds over three years. Its accounts show why a toll dollar is more than a payment on an old road bond.

• The school-tax dollar Austin keeps

  Austin ISD booked $1.51 billion in maintenance-and-operations taxes in FY2025. Recapture sent just over half away. The remaining budget has its own pressures.

• Five cities, two balance sheets

  Five Texas cities report positive General Fund balances and negative governmental unrestricted net position. The difference matters when deciding what a city can afford.

• Divisibility patterns and the size of a witness

  A superexponential lower bound for the maximum, an exact family rate, and a universal upper bound. The exact growth remains open.

• Two requested recurrence proofs

  An alternating matrix inverse, a formal binomial transform, and two explicit proof requests.

• Every ordering, from zeros and ones

  One fixed set of inputs. Every possible ordering. A constructive proof with an exact, interactive example.

• Flock: the renewal has to pay for the camera

  One annual fee bundles equipment and continuing service. I model the cash a deployment must recover and the renewal risk that can interrupt it.

• ElevenLabs: what a finished minute is worth

  Two explicit scenarios show why approval work and verified completion matter more than a cheap generation.

• Vanta after the audit

  The recurring job is keeping evidence useful. I model post-audit retention, new ARR and the investment return at the last disclosed financing benchmark.

• Who keeps the value of legal AI?

  Harvey could own more of the legal workflow. I test how recovered time becomes customer profit and what the $15.5 billion financing benchmark requires.

• Toast's gross-profit engine

  I like the workflow and the operating leverage. A full seasonal-year model puts the entry hurdle well below the dated market quote.

• Sell the review before you scale the app

  A growth channel commits you to a price, a customer and a workload. I model an assisted launch, then test the assumptions that make it win.

• Anduril is building a prime. The price assumes a platform.

  A cash-flow valuation of the factories, software economics, and financing required to turn defense demand into shareholder returns.

• Kalshi’s $40 billion question

  The market can grow faster than the fees Kalshi keeps. A valuation built around distribution, sports access, and the cash left for shareholders.

• Hyperliquid can win without HYPE being cheap.

  A $30 central case can still assume an excellent business. The difference is what reaches the token, how long growth lasts, and how many claims share it.

• Expedia: the network beyond the booking.

  A long on consumer efficiency and partner distribution, with advance travel cash removed from the valuation anchor.

• Crocs: a cash engine with a brand test.

  A long on the profitable footwear franchise, with cash-funded repurchases and an explicit tax-payment stress.

• PayPal: a modest recovery at $56.

  A long on transaction-margin recovery, valued after employee equity, lending reinvestment and customer obligations.

• Adobe: paid workflows at a price I like.

  A long on recurring creative and document demand, with a cash budget for AI, acquisitions and fewer shares.

• Texas robotaxis, one trip at a time

  A city launch gets a company onto the map. The booking rules decide whether it can get you to a meeting.

• The Texas driverless truck has to finish the job

  An empty driver’s seat matters. So do the dock, the freight left behind and the bill for a recovery.

• Chegg has bought time

  Debt retirement removes a maturity. The new business has replaced about 0.6% of the quarterly revenue lost.

• Building Atlas

  A local website audit console, and the product decisions behind a finding a reviewer can explain.

• Verifying a 54-move exact search.

  How I reconstructed a three-dimensional knight path, reduced the search, and checked the result against the puzzle rules.

• Why Texas toll roads stay tolled.

  Three years of audited accounts show how system debt, new borrowing, county transfers and concession rights keep tolls in place.

• What happened to the software buyout boom?

  An audit of 25 software take-privates: what changed ownership, what returned cash, and what the public record cannot say about returns.

• The U.S. rare-earth magnet buildout.

  Much of the disclosed plant capacity remains a future project. Qualification, product mix, inputs, and yield determine how much becomes usable supply.

• An AI megawatt is not a megawatt.

  A conditional GB300 model turns facility power into installed GPUs. A grid request alone leaves the fleet unknown.

• Who owns Austin’s home-service companies?

  Three ownership chains behind familiar Austin names, followed by the 67-brand census.

• When an index decides a company matters.

  Tracking funds turn index changes into trades. Net demand, execution, and anticipation determine what reaches the market—and whether price effects last.

• How airlines borrow against loyalty programs.

  Bank payments reach the airline before many awards are redeemed. Contracts and controlled accounts determine how lenders can claim that cash.

• Where do online returns go?

  A refund settles the customer’s claim. Freight, inspection, condition, and resale eligibility decide whether retrieving the product is worth the cost.

• Who keeps the risk in hardware finance?

  CoreWeave, Waymo, Serve, Anduril and Northvolt show where the bill lands when physical capacity earns less than planned.

• Who owns West Campus student housing?

  Six properties show how student lease obligations become property cash flow—and why rent levels cannot identify an owner’s return.

• Who funds Waymo’s hardware?

  Alphabet funds the time needed to improve the fleet. The missing asset contracts and thin modeled cash coverage show why that patience matters.

• The worker that comes back too late.

  A crawler needs a durable account of ownership, attempts and evidence before it needs a faster queue.

• Can AI crawlers see JavaScript content?

  Compare the response with the browser’s page, then identify which layer contains the answer and working links.

• The old URL is still part of the release.

  A site move succeeds when old addresses reach their intended new homes and the published signals agree.

• The Shopkeeper in the Machine

  What AI-operated shops, cafés, and radio stations reveal about pricing, memory, and the human work that keeps a business running.

• The state owns the pavement. Who owns the cash flow?

  Who collects the tolls, who controls the debt, and who bears the loss when a Texas road changes hands.

• Beyond the Leaderboard

  A code audit and proposed workflow for testing marketing-agent improvements through traces, replay, and independent evaluation.

• Can AI assistants reach your website?

  Check the crawler’s purpose, the delivered robots rules, the HTTP response and the content it actually receives.

• One person, several public records.

  The maintenance problem behind a personal site, a résumé PDF, profile markup and the biography someone else copied.

• One page, several competing identities.

  Redirects, canonical annotations, links, sitemaps and structured data should describe the same maintained document.

• Two inlinks can lead nowhere.

  Reachability begins at an intended entry. Counting links cannot tell you whether a reader can get there.

• Disallowed paths are not secrets.

  The public robots file tells cooperating crawlers where to go. Privacy begins at the server.

• How to verify AI crawlers in your server logs

  A reproducible attribution needs the request, the trusted client address, the provider’s dated ranges and the response that reached the client.

• Why Search Console Query Totals Do Not Match Page Totals

  Anonymized queries, the 1,000-row UI limit, top-row API results and property-versus-page aggregation explain many mismatched totals.

• Texas business maps: who owns, runs and finances what

  Start with the operating question, then follow the dated records behind Austin contractors, toll roads, robotaxis and freight.

• Is your local contractor still locally owned? How to check

  Follow the name on a quote through the legal entity, operating platform and investor, keeping each link attached to a dated source.

• Download the Austin home-service ownership dataset

  The 67-brand July 2026 panel, its full 75-record intake, source ledger and rules for joining later observations.

• Which Austin home-service brands share an owner?

  A dated ownership tree shows where several familiar names converge on the same platform.

• Should you block GPTBot? What changes, and what stays open?

  A decision guide for publishers who want to choose what OpenAI can collect without accidentally closing their search route.

• How can you check whether your website blocks ChatGPT, Claude or Perplexity?

  A repeatable access check starts with the exact URL, follows the delivered policy and response, and leaves a place for results you cannot yet verify.

• Is Cloudflare or your web host blocking AI crawlers?

  Trace the denied request to its rule before changing the policy you actually wanted.

• Does Google-Extended block AI Overviews, or do you need the Search Console control?

  Google now documents a Search generative AI control. It belongs beside the crawler and snippet rules, with a different job from each.

• Do Austin home-service websites allow AI crawlers?

  Most tested homepage paths were allowed by the saved robots rules. Policy and a delivered page still produced different results.

• Which robots.txt rules should you use for AI crawlers? Four complete examples

  Choose broad access, search without training, a documentation-only scope, or a decline for the listed AI agents—then test the delivered file.

• Do AI crawlers respect robots.txt? What the evidence shows

  Published controlled studies find variation. This site’s passive observations cannot supply a compliance rate.

• What 90 days of request observations show on this website

  49,152 accepted events, three observation classes and no independently verified crawler identities.

• Does anyone fetch llms.txt? What this site recorded in 90 days

  249 observed requests establish interest in a path, while reading and downstream benefit remain unmeasured.

• Before a React/Vite route goes public

  A working client-side route is only the first check. Inspect what a direct request receives, what hydration changes, and what the production host does with a missing page.

• When an AI tool call times out, did the action happen?

  A lost response leaves an operation uncertain. Durable identity, atomic receipts and a current-state check give the application a defensible recovery path.

Résumé · About Sulayman · Contact