Build1 distinct publisher3 min readPublished
A dev.to guide puts a bundle budget, a provider rate limit and a planner threshold in one frame. Run its own numbers out and the retry loop, not the invoice, is the thing that gets you.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
Each of these ceilings is enforced at an instant, while every number teams use to plan for them is an average over a window. A token bucket refills at a fixed rate up to a burst capacity, so a client can burst to capacity at once and still never sustain more than the refill [11]. Twenty requests a minute is the mean for a client that does 800 during a deploy [5]. GitHub's 5,000 an hour is 83.3 a minute, which is the figure a burst actually meets [18]. The planner has the same property in a different unit: selectivity is rows returned over rows total [15], and what the planner is handed is a bound value, not a column. A `status` column with two values averages to 0.5, while `active` is 0.95 and `pending` is 0.05 [6].
The retry arithmetic is where averaging gets expensive. In the guide's worked case, 800 requests a minute against a 500 limit with a 30 percent retry fraction gives 160 percent utilisation, 300 throttled per minute and 890 effective requests a minute; two minutes of that is priced at $3.56, or $76k a month sustained [12]. Carry the sustained figure out and it is $76,896 over a 43,200-minute month [2]. The bucket only admits 500 a minute, which is $43,200 of traffic in that same month, so about $33,700 of the headline is requests the provider refused [3]. The ceiling being modelled is on admitted throughput, not on spend, and the queue the guide prescribes (p-limit at a concurrency of 10) is the only thing that moves it, along with batching where the provider allows N calls to become 1 [13][14].
The database side prices out the same way. For `status = 'active'` on 100,000 rows with 95,000 active, the index scan costs 380k against a sequential scan's 100k, so the planner takes the scan; for a single-row `email` lookup it is 4 against 100k [16]. That is an index path 3.8 times more expensive than not using the index at all [5], for an index someone almost certainly added on purpose.
On the frontend the distinction is visible in the diff. lodash imported whole is 71.0 kB parsed and 24.4 kB gzipped, 122 ms of transfer on a 1.6 Mbps 3G link and 22 ms on 9 Mbps 4G [7]; `lodash/get` is 7.2 kB parsed, 2.9 kB gzipped, 15 ms on 3G [8], which is 8.4 times smaller on the line that matters [6]. The saving only exists if the bundler can drop the rest, and webpack tree-shakes `lodash-es` and named ESM imports, not the default `lodash` import [9]. A reviewer can check the import form by reading the patch; the gzip number needs a build, which is the argument for a CI budget of 130 kB per route and pasting the computed budget percentage into the PR description [10]. Swapping moment for date-fns is another 14 kB and 70 ms on 3G [20].
Two caveats on the thresholds themselves. The browser figure is where Lighthouse begins penalising Time to Interactive [1], a scoring penalty rather than a hard failure, which is exactly why it functions as a budget nobody notices breaking. And this is one guide, with figures reproducible through its own three calculators [19], so the constants are worth re-deriving against your provider tier and your planner before they become a merge gate.
Ranked by verification strength, evidence, and original report placement.
Cited provider rate limits: 5,000 requests per hour on GitHub, 500 per minute on OpenAI Tier 1, 100 writes per second on Stripe.
An index with 0.5 selectivity will be ignored by the planner and the query will sequentially scan 50,000 rows.
Exceed any of the three budgets and the user pays: slower paint, 429 Too Many Requests, or a query that times out.
An API that averages 20 requests per minute is 800 per minute during a burst deploy; the average hides the spike.
A table where status has two values has selectivity 0.5 on average, but 0.95 for 'active' and 0.05 for 'pending'.
lodash imported whole is 71.0 kB parsed, 24.4 kB gzip and 20.1 kB brotli; on a 1.6 Mbps 3G link that is 122 ms of transfer before first paint, and 22 ms on 9 Mbps 4G.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Internally reproducible arithmetic, externally uncorroborated
One publisher, one document. The bundle, cost and selectivity figures are self-consistent and recomputable from the guide's own formulas, which is why the derived checks hold. But every externally checkable attribution - the Lighthouse 244 kB penalty point, the 5-20 percent planner flip band, the per-provider limits, the planner cost figures of 380k versus 100k - is asserted without documentation, measurement output or a second source, and the cluster contains no telemetry, benchmark run or EXPLAIN transcript.
No adoption signal in cluster
The single source is instructional. It records no release, deployment, benchmark run, usage disclosure, pricing or licence change - only worked hypotheticals such as the recharts pull-request walkthrough. There is nothing to measure adoption from, and inferring uptake from a guide's recommendations would be invention.
Monthly-cost headline overstates billable exposure
Most of the guide is calibrated and hedged. The overstatement is localised and specific: extrapolating a two-minute burst at 890 requests per minute into '$76k per month if sustained' contradicts the guide's own token-bucket statement, and roughly $33,700 of that figure prices requests the 500 per minute bucket refuses. Framing the same scenario as throttling and retry amplification rather than spend would be accurate, so the gap is positive but modest rather than systemic.
Guide markets its own bundled calculators
The article repeatedly routes readers to four first-party tools - the Bundle Size Impact, API Rate Limit & Cost, SQL Index Selectivity and Bandwidth calculators - and stakes reproducibility of every figure on them, including advising engineers to paste calculator output into PR descriptions. That is a clear promotional channel for the author's own properties. It is disclosed in plain sight rather than hidden, and no vendor sponsorship, pricing or affiliate relationship is stated anywhere in the cluster, so the pressure is moderate.
Low: single publisher, one contested figure, no adoption data
Confidence is limited by structure rather than by contradiction. One publisher supplies every claim; the arithmetic can be rechecked but not validated against reality; one headline claim conflicts with the same document; and adoption is unmeasurable. The durable, low-risk takeaways are the mechanical ones - import shape drives wire size, retries amplify throttling, selectivity governs index use - which hold regardless of the unverified threshold values.
build
Encrypted reasoning that a cheaper sibling model can open is not protected IP1 distinct publisher
build
Your agent's retry logic is reading a timeout as a fact it does not have1 distinct publisher
build
Two Next.js apps instead of one, because isAdmin is a privilege escalation waiting to happen1 distinct publisher
build
Six weeks of agent-run ops: the failures were plumbing, not the model1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 24, 2026