35,595 results · Most relevant
Adopt the mindset of caring more about character than reputation. Acting on what you value and think is right is almost always inconvenient, which is why it's valuable; violating your own ethics and trying to get away with it erodes your self-respect. The coach lives this by explaining his students' hard lives but still declaring 'they're my lads. My responsibility.'
Our instinct when apologizing is to shrink the problem so it seems less bad. The opposite mindset — 'I have impeccable honesty and integrity' — is what makes an apology feel complete: having the courage to acknowledge what went wrong to its full extent makes the apology land rather than feel like a dodge.
The last thing you'd want to do in an intimidating situation is expose your vulnerabilities, but that's exactly what the coach does: 'I'll make amends, but just leave me lads alone.' Adopting the mindset of going first in humanizing the interaction — openly sharing what you genuinely care about — makes your motives legible, makes you more trustworthy, and gives people a chance to connect with you even mid-conflict.
What's consistent across SparkToro-style raises is no board of directors and no real investor oversight; investors' only power is veto over future fundraising or a sale that doesn't return their money if unpaid, and Rand has never gotten an unsolicited 'how's it going' check-in.
described
Under the SparkToro-style model, making more money requires paying dividends pro rata; founders can instead raise another round to keep investing, and on a sale investors and founders get pro rata ownership after payback.
described
Only 5% of venture funds over a 10–15 year period return more than 3x, which is essentially what would be needed to statistically beat the S&P 500. The speaker argues an individual without tens or hundreds of millions of dollars cannot place an appropriate number of bets to get venture outcomes that would statistically match or beat the S&P 500, and that many people who technically qualify as accredited investors could not place the statistically relevant ~500 bets on startups needed to get a venture-style outcome. His recommendation: just buy an index fund instead of investing in venture funds.
The guest offers rules of thumb: rent should be around 10% of revenue — a really good number, creeping to 15–20% depending on location (a Chadstone would push toward 20%) — and wages somewhere between 12% and 15% of revenue. Less premium locations should target closer to 10%.
Rob Walling says bootstrapping SaaS is already doing it on hard mode, and attempting it nights and weekends in an industry not conducive to that model is harder still. He does not see a path ahead for bootstrapping a 9-to-5-dependent SaaS nights and weekends in GovTech.
Rob recommends talking to other founders who have sold into the same space (both successes and failures) and to sales and marketing specialists at slightly larger but not huge SaaS companies in the same space, reached via LinkedIn.
Most VPNs are just entry and exit points: they lease a server from a landlord, set up their software on it, and your internet requests come from that node. VPN networks are real-time adjustable and API-driven, meaning an agent could use APIs to build an entire network and use VPN protocols as a way to get into that area. If a VPN is free, you are the product; you're being eaten alive. On any app store you'll probably find 50 VPNs that are all free abroad. VPN policies claiming 'no logs' raise the question of whose servers those are: are they owned by the big brand VPN or leased, and if leased, from whom, and what are those people doing with the data. There are creepy VPNs out there.
Kent C. Dodds endorses 'Tell me about the last time this happened' as a better validation question, alongside Mom Test-style questions like 'When was the last time you were stuck?' and 'How's your experience been while waiting?' — questions about actual experience rather than evaluation of an idea.
When validating an idea, don't ask users future or hypothetical questions like 'Would you pay?' or 'How much would you pay?' — nobody can tell the future, and such questions produce unreliable answers. They also turn into 'complimenting': the person just says yes to everything to get out of the conversation. Instead, don't talk about the solution at this stage; start with what the user did before — what is the past. A better validation question is 'What did you do instead?' and 'What did it cost you to do that?'
Jim Clark describes Docker's sandbox approach: a gateway endpoint is placed into each sandbox and funnels all MCP traffic. Every agent harness should have a single internal MCP gateway URL (e.g., mcpgateway.docker.internal) through which all resources and tool calls move. The gateway manages which tools, resources, and prompts are given to each sandbox, and the sandbox configuration controls what the agent can do inside that sandbox with things like MCP.
Kent C. Dodds describes the Kano model as a mental model for prioritizing software changes, with five categories of which two should never appear in your code base or product at all. Basic needs: nobody will be satisfied until they're fully implemented, no matter how much you've implemented. Performance needs: once you hit a basic level people are okay, but continuing to deliver on them makes people happier and happier — though there's probably a limit. Delighters: people aren't even expecting these, so even barely implemented they generate a lot of satisfaction, and you can continue to add delighters until they eventually cross over into performance needs.
partial (the limit on performance needs is hedged with 'probably')
Kent C. Dodds explains that not all feature requests create value in the same way, introducing the Kano model — a mental model from a 1984 paper about attractive quality and must-be quality. There are four quadrants on two axes (implemented and satisfied) and five categories, two of which should never appear in your codebase or product at all. The three categories that appear on the graph are delighters, performance needs, and basic needs. The graph has no units, so the specifics of where things fall are a bit hand-wavy.
Gemma 4 models are Apache 2 licensed, meaning users can download them, use them in their company, fine-tune them, and expand on them however is most useful.
When coming up with output choices, keep them simple — binary is best, two clear labels. With a numeric scale you would have to write a whole bunch of rules for the LLM to know the difference between a three and a four; it doesn't know that automatically. Give the judge pass-fail and it will be pretty good at deciding what is a pass and what is a fail. If you really need more categories, you can use a partial result, e.g., a 0.5 score for a half-assed job, but by default go for zero or one.
Include edge cases in your test data: non-existent tickers, multi-part questions, and jailbreak attempts. Edge cases might be 1% of your traffic, but they are the 1% that ends up being a PR disaster in the press.
After categorizing failures, sum them up into a table of root cause frequency — e.g., looks good, possible hallucination, reasoning gap, unverifiable data, missing recommendation. It's tempting to fix the most frequent failure first, but you should balance between frequency and severity instead. If one time in 100 the agent gives the user instructions on how to make a bomb instead of a financial report, you fix that one first because it's the most severe possible failure — multiply severity by frequency to decide what to fix first.
LLM judges exhibit position bias (favoring whichever option comes first, though some models prefer the last), length bias (scoring longer responses higher even when extra length is filler), confidence bias (believing wrong responses stated confidently), and self-preference bias (models tend to like their own output when generating and judging with the same model).
Choose sources → to get a library of your own and a daily brief.