Tavio | Second Mind (@Secondmindsys) header
Tavio | Second Mind (@Secondmindsys) profile picture

Built an AI reasoning harness that will break your ideas down to first principles, reveals latent intelligence, and use it to advance frameworks and workflows.

Joined October 2023
Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)Photo by Tavio | Second Mind (@Secondmindsys)
What part of your AI workflow or system do you keep correcting, working around, or trying to make more reliable? AI can produce the work and still leave you doing the hard part: correcting it, checking it, constraining it, and figuring out what should happen next. That’s the gap Second Mind Systems works on. Maybe an agent goes beyond scope. Maybe instructions don’t hold consistently. Maybe the workflow behaves differently from run to run. Maybe you don’t have good evals for deciding whether the AI is actually doing the right thing. We work across: • AI workflow diagnostics • Agent reliability audits • Eval design • Behavior and control specifications • Workflow implementation • Strategy development • Governed AI systems Right now, I’m opening the first 10 Behavior Profile Lab engagements as a bounded way to work with us. Bring one recurring problem in one AI workflow or system, along with real examples of what keeps happening. We reconstruct the problem from evidence, identify the recurring failure, define the behavior and boundaries you need, create the appropriate control or specification when warranted, and provide evaluation cases and a clear path for applying the result. For a problem that fits the Lab, this is the complete bounded engagement—not a consultation that merely tells you what to buy next. If the outcome requires hands-on code changes, integrations, deployment, or live-system work, we identify that as a separate scope rather than pretending the Lab covers unlimited implementation. Founding Lab offer: $200 for one qualifying bounded engagement. First 10 only. Public proof: Governed AI Systems Portfolio github.com/Secondmindsystems… PBM Contract Analysis I analyzed Mark Cuban’s open-source PBM contract and built a control-design contribution around how its protections survive procurement and implementation. Mark Cuban reposted the contribution on LinkedIn. Read the contribution: linkedin.com/feed/update/urn… Mark Cuban’s Repost — LinkedIn reshare panel: linkedin.com/analytics/post/… Governed Change Demo github.com/Secondmindsystems… Connect on LinkedIn: linkedin.com/in/tavio-lawren… If there’s something in your AI workflow or system you keep correcting, working around, or trying to make more reliable, DM me or email [email protected] with a short description of what keeps happening. Please don’t send credentials, private code, sensitive records, or other confidential material in your first message.
1
7
1,583
Replying to @jonzzy
It’s a rat that got demo shouted.
1
Replying to @VictorTaelin
I think the key is the learning and reasoning process itself. Say several proof attempts reveal the same missing prerequisite. That becomes the next problem to solve, and where additional compute should be allocated. If it’s proven, the other approaches can build on it. If it’s disproven, they need to change direction. The underlying structure is layered learning guided by causal dependencies. Each round should change what the next round knows, needs, and can investigate. The next reasoning step should be determined by what the previous steps actually established. Compute gets allocated to the unresolved prerequisites that unlock further progress. That’s how additional compute becomes cumulative progress rather than just more attempts.
6
Replying to @mcuban
Hey Mark, there's research supporting the connection you're making. A March 2026 regional NY Fed survey found firms facing rising health insurance costs gave average raises of 3.8%, but said they would have given 4.7% without those increases. libertystreeteconomics.newyo… A separate NBER working paper using hospital merger data estimated that a 1% increase in healthcare prices reduced both payroll and employment at employers outside healthcare by roughly 0.4% each. nber.org/papers/w32613 These support the connection between healthcare costs, wages and employment. Neither establishes that healthcare costs outweigh every other factor or proves your proposed 50% scenario. Your wage hypothesis gives us a specific calculation to investigate: what would redirecting half the increase in benefits spending add to wage growth, and would that put it above inflation? Use the same workers and time period, distinguish employer spending from employee contributions, and avoid counting the same dollars twice. Lower employee deductions increase take home pay, while money added to salaries increases gross wages. That would put a number on the opportunity you're describing: how much of rising healthcare spending could instead become stronger purchasing power for workers.
1
1
81
@grok can you substantiate those ranges? Link the source and valuation date for each European club estimate, then show the named before-and-after club valuations supporting the 20–50% relegation decline. Keep revenue losses and squad-value changes separate. If comparable club valuations aren’t available, withdraw the percentage and show what the evidence actually establishes. Explain in plain English.
1
135
Replying to @mcuban @grok
Mark, here's how to use Grok to surface the latent intelligence behind your original questions. @grok answer Mark’s comparison, then investigate that distinction “Find the five lowest valued clubs in MLS, the Premier League and La Liga. Cite valuation dates and sources. Separate whole club enterprise value, equity value and squad value. If coverage is incomplete, show the gaps. For relegated clubs, find comparable valuations or transactions before and after the drop. Separate valuation changes from revenue declines, and distinguish risk already priced in from losses after relegation was confirmed. Account for debt, stadium ownership, wage commitments, relegation payments and the prospect of promotion. Report a typical percentage only if comparable evidence supports one. Then compare similarly priced MLS and European clubs: what does the buyer acquire, what income is at risk, and what could recovery require? Explain what the evidence establishes and what remains uncertain in plain English.
3
3,218
Replying to @kimmonismus
I’m pretty sure this is what everyone wants to see. The same thing happened with Altman.
61
Replying to @garrytan
Garry, the two attempts per task give you a way to decide what harder* should mean for the next benchmark. Use the disagreements to find out what each model handles differently. Give this to your eval agent. “Use the raw results and task descriptions for the Opus 5.5 and Fable 5.2 runs. For every task, show how many attempts each model passed out of two. Separate shared wins, shared failures, wins that appeared once, and tasks one model passed twice while the other failed twice. Flag environment failures separately. If you only have the chart, ask for the per-attempt results. Review the tasks and traces with model names hidden. Describe what made each task difficult and what actually stopped the failed attempts. Then reveal the model names and look for a pattern that could explain where either model had a repeatable edge. Turn the strongest pattern into a prediction about fresh tasks. Specify which feature of the task should produce the gap, pick new tasks with and without that feature, and write down what result would disprove the prediction before running them. Keep the harness, grading, and attempt count fixed. Report solve rate and cost per solve for each group. Finish with one workload where the evidence supports a default, one where the premium might pay off, and what you’d still need to measure to make that call.”
3
280
Replying to @Warcraft
A baseline racial moving as fast as mounts and faster than druid ferals, Druid travel forms, ghost wolf, aspect of cheetah, some rogue sprints - not balanced. 30% was good enough.
17
1
40
5,150
Replying to @baldwinbuilds
I have like 4 alts lvs 16-19. Safe to say I won’t be making it lol.
1
2
678
Replying to @mcuban
I’d turn “elsewhere” into a question every employer can ask at renewal. Where did the savings land? Pick one specialty drug. Compare the same drug and quantity under the current plan and a viable alternative. Show what the employer pays after fees and rebates, what the patient pays out of pocket, and what the vendor and its affiliates receive where those payments can be verified. Check whether patients can still get the drug just as reliably and quickly. If one bill drops and another rises, show who got the savings and who picked up the cost. If the combined bill drops and patients can still get their medication on the same terms, that’s evidence of a better deal.
83
Replying to @ChrisGPT
Ilya, Andrew, and Demis.
62
Replying to @hwchase17
Pretty much intent engineering.
2
317
Replying to @tiffplzz
Terrible racials is the only thing I can think of. They are aesthetically awesome.
1
118
Replying to @willdepue
Mythos already picked 68 when it reasoned that matching the key with a caveat might preserve transparency through training. I’d test two things separately. First, hold the model fixed and vary whether it expects the grader to favor the wrong answer. Does that change what it submits or discloses? Then start from the same checkpoint and train matched copies with graders that score the same final answers but differ in whether they reward a well-supported caveat. On new tasks with flawed keys and no grader cues, does one copy flag the errors more often? The first test measures whether reward expectations change the answer. The second measures whether rewarding the caveat changes what survives training. anthropic.com/document/claud…
1
521
Replying to @kimmonismus
Finally. Ngl, more excited about this than Opus.
1
334